Structure Determination and Prediction of Zeolites -- a Combined Study by Electron Diffraction, Powder X-Ray Diffraction and Database Mining

Home , Chabazite, Gmelinite

Structure Determination and P r e diction of Z e o l i t e s - A Combined Study by Electron Diffraction, Powder X - Ray Diffraction and Database M i n i n g

Peng Guo

Structure Determination and Prediction of Zeolites -- A Combined Study by Electron Diffraction, Powder X-Ray Diffraction and Database Mining

Peng Guo 郭鹏

Doctoral Thesis 2016

Department of Materials and Environmental Chemistry Arrhenius Laboratory, Stockholm University SE-106 91 Stockholm, Sweden

Cover: An old zeolite ZSM-25 is woke up by an alarm

Faculty opponent: Prof. Christine Kirschhock Center for Surface Chemistry and Catalysis KU Leuven Belgium

Evaluation committee: Dr. Johanne Mouzon Department of Civil, Environmental and Natural Resources Engineering Luleå University of Technology

Prof. Vadim Kassler Department of Chemistry and Biotechnology Swedish University of Agricultural Sciences

Dr. German Salazar Alvarez Department of Materials and Environmental Chemistry Stockholm University

Substitute: Dr. Mårten Ahlquist Theoretical Chemistry and Biology KTH Royal Institute of Technology

©Peng Guo, Stockholm University 2016 ISBN 978-91-7649-384-7

Printed by Holmbergs, Malmö 2016 Distributor: Department of Materials and Environmental Chemistry

A shrewd and ambitious life needs no explanation.

---Yong-hao Luo (罗永浩)

To my family

Abstract

Zeolites are crystalline microporous aluminosilicates with well-defined cavities or channels of molecular dimensions. They are widely used for applications such as gas adsorption, gas storage, ion exchange and catalysis. The size of the pore opening allows zeolites to be categorized into small, medium, large and extra-large pore zeolites. A typical zeolite is the small pore sili- coaluminophosphate SAPO-34, which is an important catalyst in the MTO (methanol-to-olefin) process. The properties of zeolite catalysts are determined mainly by their structures, and it is therefore important to know the structures of these materials to understand their properties and explore new applications. Single crystal X-ray diffraction has been the main technique used to determine the structures of unknown crystalline materials such as zeolites. This technique, however, can be used only if crystals larger than several micro- metres are available. Powder X-ray diffraction (PXRD) is an alternative technique to determine the structures if only small crystals are available. However, peak overlap, poor crystallinity and the presence of impurities hinder the solution of structures from PXRD data. Electron crystallography can overcome these problems. We have developed a new method, which we have called “rotation electron diffraction” (RED), for the automated collection and processing of three-dimensional electron diffraction data. This thesis describes how the RED method has been applied to determine the structures of several zeolites and zeolite-related materials. These include two interlayer expanded silicates (COE-3 and COE-4), a new layered zeolitic fluoroaluminophosphate (EMM-9), a new borosilicate (EMM-26), and an aluminosilicate (ZSM-25). We have developed a new approach based on strong reflections, and used it to determine the structure of ZSM-25, and to predict the structures of a series of complex zeolites in the RHO family. We propose a new structural principle that describes a series of structurally related zeolites known as “embedded isoreticular zeolite structures”, which have expanding unit cells. The thesis also summarizes several common structural features of zeolites in the Database of Zeolite Structures.

Key words: zeolites, rotation electron diffraction, structure determination, structure prediction, strong reflections approach

List of papers

Paper I: Ab initio structure determination of interlayer expanded zeolites by single crystal rotation electron diffraction

Peng Guo, Leifeng Liu, Yifeng Yun, Jie Su, Wei Wan, Hermann Gies, Hai- yan Zhang, Feng-Shou Xiao and Xiaodong Zou. Dalton Trans., 2014, 43, 10593–10601.

Scientific contributions: I conducted the TEM work, carried out the structure solution, made the Rietveld refinement, wrote and corrected the manuscript.

Paper II: Synthesis and structure determination of a layered zeolitic fluoroaluminophosphate and its transformation to a three-dimensional zeolite framework

Peng Guo, Guang Cao, Mobae Afeworki, YifengYun, Junliang Sun, Jie Su, Wei Wan and Xiaodong Zou. In manuscript

Scientific contributions: I conducted the TEM work, carried out the structure solution, made the Rietveld refinement, and wrote the manuscript.

Paper III: EMM-26: a two-dimensional medium pore borosilicate zeolite with 10×10 ring channels solved by rotation electron diffraction

Peng Guo, Karl Strohmaier, Hilda Vroman, Mobae Afeworki, Peter I. Ra- vikovitch, Charanjit S. Paur, Junliang Sun, Allen Burton and Xiaodong Zou. In manuscript

Scientific contributions: I conducted the TEM work, carried out the structure solution, made the Rietveld refinement, and wrote the manuscript.

Paper IV: A zeolite family with expanding structural complexity and embedded isoreticular structures

Peng Guo#, Jiho Shin#, Alex G. Greenaway, Jung Gi Min, Jie Su, Hyun June Choi, Leifeng Liu, Paul A. Cox, Suk Bong Hong, Paul A. Wright and Xiaodong Zou. Nature, 2015, 524, 74–78. (# Equal contribution)

Scientific contributions: I conducted the TEM work, carried out the structure solution and structure prediction work, made the Rietveld refinements, wrote and corrected the manuscript.

Paper V: Targeted Synthesis of Two Super-Complex Zeolites with Embedded Isoreticular Structures

Jiho Shin, Hongyi Xu, Seungwan Seo, Peng Guo, Jung Gi Min,Jung Cho, Paul A. Wright, Xiaodong Zou and Suk Bong Hong. Angew. Chem. Int. Ed., 2016, DOI: 10.1002/anie.201510726.

Scientific contributions: I predicted structural models and wrote the structure prediction part of the manuscript.

Paper VI: On the relationship between unit cells and channel systems in high silica zeolites with the "butterfly" projection

Peng Guo, Wei Wan, Lynne McCusker, Christian Baerlocher and Xiaodong Zou. Z. Kristallogr., 2015, 230, 5, 301–309.

Scientific contributions: I identified the related structures, analyzed them, wrote and corrected the manuscript.

Papers not included in the thesis

Paper VII: The Use of Porous Palladium(II)-polyimine in Cooperatively- catalyzed Highly Enantioselective Cascade Transformations

Chao Xu, Luca Deiana, Samson Afewerki, Celia Incerti-Pradillos, Oscar Córdova, Peng Guo, Armando Córdova, and Niklas Hedin. Adv. Synth. Catal., 2015, 357, 2150–2156 Scientific contributions: I conducted the TEM work.

Paper VIII:

Fabrication of novel g-C3N4/nanocage ZnS composites with enhanced photocatalytic activities under visible light irradiation

Jing Wang, Peng Guo, Qiangsheng Guo, Pär G. Jönsson and Zhe Zhao. CrystEngComm, 2014, 16, 4485–4492.

Scientific contributions: I conducted the TEM work and corrected the manuscript.

Paper IX: Visible light-driven g-C3N4/m-Ag2Mo2O7 composite photocatalysts: synthesis, enhanced activity and photocatalytic mechanism

Jing Wang, Peng Guo, Maofeng Dou, Jing Wang, Yajuan Cheng, Pär G. Jönsson and Zhe Zhao. RSC Adv., 2014, 4, 51008–51015.

Scientific contributions: I conducted the TEM work and corrected the manuscript.

Paper X: Rapid sintering of silicon nitride foams decorated with one-dimensional nanostructures by intense thermal radiation

Duan Li, Elisângela Guzi de Moraes, Peng Guo, Ji Zou, Junzhan Zhang, Paolo Colombo and Zhijian Shen. Sci. Technol. Adv. Mater., 2014, 15, 045003–04509.

Scientific contributions: I conducted the TEM work and corrected the manuscript.

Paper XI: One-pot Synthesis of Metal-Organic Frameworks with Encapsulated Target Molecules and Their Applications for Controlled Drug Delivery

Haoquan Zheng, Yuning Zhang, Leifeng Liu, Wei Wan, Peng Guo, Andreas M. Nyström and Xiaodong Zou. J. Am. Chem. Soc., 2016, 138, 962–968

Scientific contributions: Haoquan and I identified this unique material.

Paper XII: Two ligand-length-tunable interpenetrating coordination networks with stable Zn2 unit as three-connected uninode and supramolecular topolo- gies

Guohai Xu, Jianyi Lv, Peng Guo, Zhonggao Zhou, Ziyi Dua and Yongrong Xie. CrystEngComm, 2013, 15, 4473–4482.

Scientific contributions: I did structure and topology analysis and corrected the manuscript.

Contents

1. Introduction ...... 15 2. Zeolites ...... 19 2.1 Zeolite structure ...... 19 2.1.1 Building units ...... 20 2.1.2 Pore system ...... 23 2.1.3 Non-framework species ...... 23 2.2 Properties of zeolites ...... 25 2.2.1 Small pore zeolites ...... 25 2.2.2 Medium pore zeolites ...... 25 2.2.3 Large pore zeolites ...... 26 3. Structure determination of zeolites ...... 27 3.1 Basic crystallography ...... 27 3.1.1 Crystals and crystallographic symmetry in real space ...... 27 3.1.2 Reciprocal space ...... 29 3.1.3 Structure factors ...... 30 3.1.4 Structure determination by diffraction ...... 32 3.1.5 Algorithms for the structure determination ...... 34 3.2 Structure determination of zeolites ...... 37 3.2.1 Single crystal X-ray diffraction (SCXRD) ...... 37 3.2.2 Powder X-ray diffraction (PXRD) ...... 37 3.2.3 FOCUS ...... 38 3.2.4 Rotation electron diffraction (RED) ...... 39 3.2.5 HRTEM ...... 40 3.2.6 Model building ...... 42 3.3 Rietveld refinement ...... 42 4. Structure determination of zeolites and zeolite-related materials by rotation electron diffraction (RED) ...... 45 4.1 COE-3 and COE-4 (Paper I) ...... 45 4.2 EMM-9 (Paper II) ...... 49 4.3 EMM-26 (Paper III) ...... 54 4.4 Conclusions ...... 58 5. Unravelling the structural coding of the RHO zeolite family ...... 59

5.1 Structure determination of ZSM-25 (Paper IV) ...... 59 5.2 Structure predictions of PST-20 (RHO-G5) and PST-25 (RHO-G6) . 67 5.3 Structure predictions of PST-26 (RHO-G7) and PST-28 (RHO-G8) (Paper V) ...... 68 5.4 Conclusions ...... 70 6. Database mining of zeolite structures ...... 72 6.1 Characteristic structural information hinted by the unit cell dimensions ...... 72 6.1.1 5 Å ...... 72 6.1.2 7.5 Å ...... 74 6.1.3 10 Å ...... 77 6.1.4 12.7Å ...... 79 6.1.5 14 Å ...... 80 6.1.6 20 Å ...... 81 6.2 The ABC-6 family ...... 81 6.3 The butterfly family (Paper VI) ...... 85 6.4 Conclusions ...... 92 7. Sammanfattning ...... 93 8. Future perspective ...... 95 9. Acknowledgements ...... 97 10. References ...... 99

Abbreviations

IUPAC International Union of Pure and Applied Chemistry MOF Metal-organic framework COF Covalent organic Framework FTC Framework type code IZA International Zeolite Association 3D Three-dimensional 2D Two-dimensional SBU Secondary building unit CBU Composite building unit MTO Methanol to olefin SCR Selective catalytic reduction SEM Scanning electron microscopy OSDA Organic structure directing agent sod Sodalite Fhkl Structure factor HRTEM High resolution transmission electron microscopy Ehkl Normalized structure factor FOM Figure of merit SCXRD Single crystal X-ray diffraction PXRD Powder X-ray diffraction RED Rotation electron diffraction SAED Selected area electron diffraction FWHM Full-width at half-maximum height COE International Network of Centers of Ex- cellence IEZ Interlayer expanded zeolite AlPO Aluminophosphate TEA+ Tetraethylammonium TPA+ Tetrapropylammonium CTF Contrast transfer function GOF Goodness of fit EDTA Diethylenetriamine CIF Crystallographic information file

1. Introduction

Porous materials are promising and important materials distinct from traditional dense materials such as Au, TiO2 and CdS. Porous materials are widely used in the adsorption, catalysis, gas separation and purification, and energy storage. The International Union of Pure and Applied Chemistry (IU- PAC) (1) has categorized porous materials into three types based on the size of their pores: microporous (with a pore size smaller than 2 nm), mesoporous (2-50 nm) and macroporous (larger than 50 nm). Five important classes of porous materials have recently been reviewed by Prof. Andrew I. Cooper at the University of Liverpool (2) (Figure 1.1). Zeolites are considered to be “traditional” porous materials, while the other classes reviewed by Cooper (metal-organic frameworks (MOFs), covalent organic frameworks (COFs), porous organic polymers and porous molecular solids) have been developed more recently, during the past twenty years. These new porous materials can be tailored or given specific functions very easily by the elaborate design of organic motifs. Yaghi, for example, has shown MOF-74-XI, which belongs to a series of MOF-74 structures, can be given a pore size in the mesoporous range (9.8 nm) by expanding the organic linkers (3). Another example is from Mircea Dinca’s research group at the Massachusetts Institute of Tech- nology, who created a series of electroactive thiophene COFs. One of these is an unusual charge-transfer complex with tetracyanoquinodimethane (TCNQ) (4). However, comprehensive parameters, such as selectivity, kinetics, mechanical properties and stability, are more important for large-scale industrial applications. Since “traditional” zeolites have suitable properties in these respects, they are widely used and have not yet been replaced by these promising new porous materials (Section 2.2). The first natural zeolite, stilbite (whose framework type code (FTC) is “STI”), was discovered 260 years ago by a Swedish mineralogist, Axel Fredrik Cronstedt (5), while the first synthetic zeolite (levynite, FTC: LEV) (6) was reported in 1862 by St. Claire Deville, who mimicked the conditions in which natural zeolites formed. In 1948, Barrer, a pioneer in the systematic synthesis of zeolites, obtained the first unknown zeolite (ZK-5, FTC: KFI) (7, 8), where “unknown” denotes that no natural counterpart was known at that time. Another breakthrough in the synthesis of zeolites came in 1961, when Barrer and Denny utilized quaternary ammonium cations to synthesize zeolites. The widely-used ZSM-5 (9) and zeolite beta (10) were synthesized

by this method. This approach to synthesis has remained popular and is an efficient method for synthesizing new zeolites (11, 12).

One

Figure 1.1 Classes of porous materials and selected functions. Reprinted with permission from Ref. 2. Copyright © 2015, the American Association for the Advancement of Science. Zeolite scientists have long pursued large-pore zeolites that can accom- modate large molecules and facilitate their mass transport. However, small-pore zeolites have recently come into focus. These zeolites have shown to be useful in the methanol-to-olefin (MTO) process (13), gas separation (14–17) and selective catalytic reduction (SCR) (18). Zeolite researchers have traditionally showed their passion by synthesising zeolites with complex three-dimensional frameworks. Now, however, one of the hot research topics has become the post-synthesis of two-dimensional zeolite- related materials. Two-dimensional ferrierite layers are one example: they can be obtained by etching double 4-rings in a known germanosilicate, IM- 12 (FTC: UTL). At least seven new zeolites have been synthesized through further careful manipulation of these 2D layers by fine tuning the pH, adding extra sources or silica, or using other organic templates (19–22). It is necessary to know the atomic structure of a zeolite to understand its properties and to perform precise post-synthesis and modification. The dimensions of the crystallographic unit cell, the crystal space group and the

positions of atoms are all determined during structure determination. The history of the structure determination of zeolites shows that it is based on data obtained in real space (model building and high resolution transmission electron microscopy (HRTEM) imaging), reciprocal space (several diffraction techniques, including single crystal X-ray diffraction (SCXRD), powder X-ray diffraction (PXRD) and electron diffraction), and a combination of both. Refinement of a preliminary structural model against the experimental data has been used to confirm whether the model is correct. For example, Pauling and Taylor solved the structures of six zeolites (23–27) in the 1930s by combining model building with careful analysis of the crystal symmetry (unit cell dimensions and space group), which had been obtained from SCXRD and PXRD. Zeolite structures could not at that time be solved from diffraction data directly. Even today, the model building approach is very helpful. Some structural features of the zeolites studied here are summarized in Chapter 6, in order to make the structure determination of zeolite by this method much more convenient. X-ray diffraction techniques have matured, and algorithms for phasing have been developed, and thus SCXRD and PXRD have become the main tools for the ab initio solution of unknown structures including zeolites. SCXRD is limited by the availability of sufficiently large crystals (crystals of dimensions around 20 µm are needed for in-house SCXRD diffractometers), while PXRD suffers from reflection overlap, the presence of impurities, poor crystallinity, and disorder. Electron crystallography, which includes both electron diffraction and HRTEM, can overcome these problems. One recent breakthrough in electron crystallography is the development of 3D electron diffraction techniques including automated diffraction tomography in Ute Kolb’s group in Mainz (28–30) and rotation electron diffraction (RED) in our research group of (Section 3.2.4) (31, 32). This technique allows 3D electron diffraction data to be collected from nano-sized crystals. The 3D ED can be used to determine structures by employing known algorithms for phasing. The ED data, however, suffers from two main problems: dynamical effects and electron beam damage, which makes it difficult to carry out accurate refinement against ED data. Several novel structures have been solved from 3D ED data (33–42). In the work presented in this thesis, RED has been used to determine the structures of submicrometer-sized zeolites. The structures have then been refined by the Rietveld technique using PXRD data. The interaction between non-framework species and the framework has been elucidated through Rietveld refinement. In addition to the structure determination of novel zeolite structures, the thesis includes also structure prediction of new zeolites with tailored properties. Michael W. Deem et al. have constructed a database of computationally predicted zeolite-like materials using a Monte Carlo search (43). We have, in contrast, developed an approach for predicting structures based on the al-

ready-known structures of zeolites. Our approach can also link the structure prediction with the synthesis of zeolites. The main objects of the work presented here were: 1) To solve the structures of several zeolite-related materials (COE-3, COE- 4, and EMM-9) and a borosilicate zeolite EMM-26 using the RED method by direct methods. 2) To use Rietveld refinement to validate the structural models obtained from RED data and to elucidate the interactions between non-framework species and the framework. 3) To develop a new approach (which we have called “the strong reflections approach”) to identify a zeolite family, determine structures (ZSM-25, FTC: MWF), and predict structures (PST-20, PST-25, PST-26 and PST-28). In a way, this unique approach can guide the synthesis of zeolites. 4) To summarize common structural features of known zeolites. This summary may help in the structure determination of unknown zeolites, and inspire the synthesis of new zeolites with similar structural features.

2. Zeolites

Zeolites are crystalline microporous aluminosilicates with well-defined cavities and/or channels. Due to their wide applications for ion exchange, gas separation, gas storage and organic catalysis, zeolites have drawn increasing attention from both academia and industry. The term “zeolite” was originally coined in 1756 by the Swedish mineralogist Axel Fredrik Cronstedt. When he rapidly heated the mineral stilbite (FTC: STI), a large amount of steam was produced from water that had been absorbed in the mineral. Based on this, he called the material “zeolite”, which is derived from two Greek words: “zéō”, to boil and “líthos”, a stone (5). The basic crystallographic building unit of a zeolite is TO4 (T, tetrahedron), where the T atom can be Si or Al. The typical distances of Si-O, O-O and Si-Si are 1.61 Å, 2.63 Å and 3.07 Å, respectively, in the pure silica form (Figure 2.1). The TO4 tetrahedra connect with the adjacent tetrahedra through corner-sharing, generating the three- dimensional (3D) framework of a zeolite. Replacement of Si4+ with Al3+ in zeolites results in negative charges in the framework. Inorganic cations (such as Li+, Na+ and K+), organic cations (such as TPA+, tetrapropylammonium), or a mixture of both can be introduced into the channels or cavities of zeolites to balance the negative charges from the framework, making the total charge of the entire structure neutral. The chemical elements initially identified in zeolites (Al and Si) have now been extended to include B, P, Ti, V, Mn, Fe, Co, Ni, Zn, Ga and Ge. This extension has made the structures and properties of zeolites much more diverse and fascinating, and has opened an avenue to create zeolites with larger pores than conventional aluminosilicate zeolites (44).

2.1 Zeolite structure The International Zeolite Association (IZA) has approved 231 zeolite FTCs. Some common structural features are present in different zeolite structures, and these will be introduced and summarized in this section.

Figure 2.1 Typical Si-O, O-O and Si-Si distances in the pure-form silica zeolite.

2.1.1 Building units As mentioned before, the basic building units of zeolites are tetrahedra with chemical formula TO4 (T=Al, Si, P, Ge, B...). These can form a number of larger building units. Zeolite structures are usually described in terms of secondary building units (SBUs), composite building units (CBUs), chains and layers. SBUs should satisfy the following requirements (45): (1) The entire framework should be constructed based on one single unit; (2) The number of SBUs within one unit cell should be an integer; (3) The maximum number of T atoms in one SBU is 16. All SBUs are summarized in the IZA Database of Zeolite Structures (46). It is fascinating to see how the same SBUs with different connections generate a variety of zeolite structures. For example, SFO, AFR, ZON, JSN (type material: MAPO-CJ69) and OWE frameworks have the same 4-4- SBUs, which can be described as double 4-rings with one-edge disconnected (47, 48). “Head-to-tail” arrangements of 4-4- SBUs appear in the first four frameworks, while “shoulder-to-shoulder” arrangements of these SBUs show up in the OWE framework. In addition, the layers in the SFO and AFR frameworks are identical, but linked in different ways. The former is linked via inversion center, while the latter is connected via mirror symmetry, as shown in Figure 2.2. In this thesis, a 2D layered structure with the 4-4- SBUs, EMM-9, will be introduced in Chapter 4.

Figure 2.2 The complete building process from 4-4- SBUs to three couples of structurally closely related structures: SFO and AFR, JSN (type material: MAPO-CJ69) and ZON, OWE and a hypothetical structure H-CJ69. Re- printed with permission from Ref. (47). Copyright (2012) American Chemi- cal Society. Although SBUs can be used as the only building unit to describe a certain zeolite structure, it is more interesting to use composite building units (CBUs). CBUs are the building units that are frequently found in several zeolites. A zeolite structure can be built using more than one CBU. All CBUs are summarized in the IZA Database of Zeolite Structures (49). Figure 2.3 shows some common and simple CBUs. Take the CBU lta cage for example; it is composed of twelve 4-rings, eight 6-rings and six 8-rings, so the tile symbol for it can be written as [4126886]. This cage can be found in the - CLO, KFI, LTA, LTN, PAU, RHO, TSC and UFI frameworks. Some zeolite structures consist of chains. The three most common chains, the double zig-zag, double saw-tooth and double crankshaft chains are shown in Figure 2.4. The approximate periodicity is 5 Å for the double zig- zag chain, 7.5 Å for the double saw-tooth chain and 10 Å for the double crankshaft chain. It is useful to know these characteristic chain periodicities when determining the structures of unknown zeolites (Chapter 6). A nomenclature similar to that used for chains has been developed to describe 2D three-connected layers. Some zeolites can be built from a single type of layers. Neighbouring layers are related either by a simple translation or by symmetry operations, and are further connected to construct a variety of zeolite structures from the same basic layer. Consider the structure of the

layer denoted as “4·82 layer” (Figure 2.5), for example; each node is associ- ated with one 4-ring and two 8-rings, leaving the fourth connection pointing up or down. The combination of different up and down possibilities generates a variety of zeolite structures, examples of which are the GIS and ABW frameworks. Each node within the 4·82 layer is three-connected, leaving the fourth connection pointing up (highlighted in blue) or down (highlighted in yellow) (Figure 2.5). Another zeolite family with a “butterfly” layer will be presented in detail in Chapter 6.

Figure 2.3 Seven cages found in zeolites.

Figure 2.4 Three types of chains frequently observed in zeolite structures.

Figure 2.5 The up-down configurations of 4·82 layers in the GIS and ABW frameworks. “Up” and “down” modes are highlighted in blue and yellow, respectively.

2.1.2 Pore system Zeolites can be categorized into four categories, based on the numbers of TO4 tetrahedra that define the pore window: small pore (delimited by 8 TO4), medium pore (10 TO4), large pore (12 TO4) and extra-large pore (more than 12 TO4) zeolites. The diameters of the pore openings are normally approximately 3.8 Å in small pore, 5.3 Å in medium pore and 7.4 Å in large pore zeolites, respectively. The pore diameter in the extra-large pore VPI-5 material with 18-ring pores (FTC: VFI) is approximately 12.7 Å. However, pore openings may be ellipsoid or may have more complicated shapes, and thus some medium pore or large pore zeolites appear as small pore zeolites. For example, a 3D open framework borogermanate SU-16 (FTC: SOS), synthesized using diethylenetriamine (EDTA) (50), has an elliptical 12-ring opening with a small effective pore size (3.9 Å × 9.1 Å) (51). In Chapter 4, another new medium pore zeolite EMM-26 but with a small effective pore opening will be introduced. It shows effective selectivity of CO2/CH4.

2.1.3 Non-framework species Positions of inorganic cations and organic structure directing agents (OSDAs) play significant roles in determining the properties of zeolites and crystal growth mechanisms. The importance of the positions of the inorganic cations and OSDAs is described below, using Zeolite A (FTC: LTA) and SSZ-52 (FTC: SFW) as examples. The effective pore opening of Zeolite A can be tuned depending on the type of inorganic cations that occupy the pores. The pore diameter is 3 Å with K+ (known as Zeolite 3A), 4 Å with Na+ (Zeolite 4A) and 5 Å with Ca2+ (Zeolite 5A). As shown in Figure 2.6, dehydrated Zeolite 3A demonstrates

three typical positions for K+: in the middle of the 6-ring, on the edge of the 8-ring and near the 4-ring. The distribution of Na+ ions in the dehydrated Zeolite 4A is similar to that in Zeolite 3A, but the effective pore size is greater, approximately 4 Å because of the smaller size of Na+ compared to that of K+. The Ca2+ ions are located in the middle of a 6-ring of the -cage in the dehydrated Zeolite 5A, expanding the 8-ring pore size to 5 Å. The pore opening can be further tuned by incorporating more than one type of inorganic cations. For example, NaK Zeolite A, containing a mixture of Na+ + and K cations in a ratio of 83:17, shows a very high ideal CO2 (kinetic diameter: 3.3 Å) /N2 (kinetic diameter: 3.6 Å) selectivity (52). This phenome- non can be explained by the “trapdoor” mechanism (14, 15). SSZ-52 (FTC: SFW) provides an example of the importance of OSDAs. This compound was synthesized at Chevron Energy and Technology Com- pany using an unusual polycyclic quaternary ammonium cation as the OSDA (53). The structure of the zeolite was solved by high-level model building, based on the unit cell parameters of SZZ-52. Seven possible structural models were built, one of which was compatible with the experimental data. Rietveld refinement showed that there are two OSDAs in a single cavity. The pair of OSDAs with the “head-to-head” configuration (the tail has a positively charged amine) directed the arrangement of the double 6-rings surrounding them. This detailed structural information helped to understand the growth mechanism of the SSZ-52 zeolite.

Figure 2.6 Zeolite A -cage showing the locations of inorganic cations in the pore. (a) Zeolite 4A with Na+ in the pore, (b) Zeolite 3A with K+ in the pore and (c) Zeolite 5A with Ca2+ in the pore.

2.2 Properties of zeolites

2.2.1 Small pore zeolites The methanol to olefin (MTO) process (13), gas separation (14–16) and selective catalytic reduction (SCR) (18) have stimulated renewed interest in the application of small pore zeolites as molecular sieves. This section focuses on the gas separation such as CO2/N2 and CO2/CH4. Natural gas is a hydrocarbon gas mixture formed in nature, consisting primarily of methane with small amounts of carbon dioxide, nitrogen, and hydrogen sulphide. The removal of undesirable CO2 in order to upgrade natural gas was previously mainly carried out by aqueous amine scrubbing (54), which requires much energy. The separation can be carried out more simply and in a more environmentally friendly manner using small pore zeolites. For example, the effective pore size of the RHO framework is about 3.6 Å, which allows the smaller carbon dioxide molecules (with a kinetic diameter of 3.3 Å) to pass through and prevents larger molecules such as nitrogen (kinetic diameter: 3.6 Å) and methane (kinetic diameter: 3.8 Å). Furthermore, the gas selectivities of CO2/CH4 and CO2/N2 can be further enhanced using the cations-exchanged zeolite Rho (RHO). Displacements of cations located at the 8-ring sites of the Rho allow CO2 uptake to occur, which explains this promising result.

2.2.2 Medium pore zeolites ZSM-5, initially discovered in 1965 by Mobil Technology Company (9), and synthesized using tetrapropylammonium as an OSDA, is a classical medium pore zeolite with straight 10-ring channels along the b-axis and zig-zag 10- ring channels along the a-axis (55). It is widely used in many applications such as propylene production (56), gasoline octane improvement (57), methanol to hydrocarbons conversion (58) and the product selectivity of xylene (59, 60). The para-selectivity of modified ZSM-5 zeolites is an example of the product selectivity of xylene. Several products, such as ethylbenzene, sty- rene and xylene can be formed over zeolites through alkylation of toluene with methanol, which results in the insertion of the methyl group in the chain or in the ring (59, 60). The selectivity over a certain product can be tuned by controlling the acidity/basicity of the zeolites. Normally, acidic zeolite catalysts are used for the formation of xylene through the methylation of toluene. Furthermore, a high concentration of methanol will promote the formation of trimethylbenzene (which is formed by a bimolecular mechanism), while a low concentration of methanol will eliminate the formation of trimethylbenzene and facilitate the formation of xylene. B- and P-modified ZSM-5 have

higher para-selectivities, because the guest ions B and P decrease the free diameter of the catalyst, leading to the rapid release of p-isomers.

2.2.3 Large pore zeolites Zeolite Y, with framework type code FAU, crystallizes in the cubic crystal system with a = 24.74 Å. The composite building units in Zeolite Y are double 6-rings and sodalite (sod) cages. Each sod cage connects with four double 6-rings (Figure 2.7a), while each double 6-ring links to two sod cages, generating a 3D channel system with 12-ring pore openings (with a pore size of 7.4 Å) (Figure 2.7c). Zeolite Y shows a promising catalytic performance in fluid cracking catalysis, which is used to convert the high-boiling hydrocarbon fractions of crude oils to more valuable gasoline and other products. This is mainly due to its unique properties: (1) high surface area and relatively large pore size; (2) strong Brønsted acidity; and (3) excellent thermal and hydrothermal stability (61).

Figure 2.7 (a) One sod cage connects with four d6rs; (b) one d6r links two sod cages; (c) a 12-ring highlighted in dark blue in the FAU framework.

3. Structure determination of zeolites

An introduction to crystallography will be given before we go on to consider solving the structures of zeolites. Crystallography is useful not only in the zeolite field, but also for unravelling the structure of any unknown crystalline material. Structural analysis based on crystallography can be done in two spaces: real space (direct space) and reciprocal space (diffraction space).

3.1 Basic crystallography

3.1.1 Crystals and crystallographic symmetry in real space The following introduction to crystallography is based mainly on two crystallography books: Phasing in Crystallography by Carmelo Giacovazzo (62), and Electron Crystallography by Xiaodong Zou, Sven Hovmöller and Peter Oleynikov (63).

Crystal According to the definition of The International Union of Crystallography (IUCr), “a material is a crystal if it has essentially sharp peaks in its diffraction pattern. The word ‘essentially’ is used to describe the situation in which most of the diffraction intensity is concentrated in relatively sharp Bragg peaks, with a small fraction in the always present diffuse scattering ”(64).

Unit cell Crystals are periodic in three dimensions. It is unnecessary to describe a crystal by the individual atom. The crystal can be described by a smallest and periodically-repeated parallelepiped, which is called “unit cell”. The unit cell should satisfy the following requirements: 1) The content and size of all the unit cells in a perfect crystal are identical. 2) The entire crystal is constructed by the edge-to-edge translation of the unit cells in 3D, without any rotation or mirror symmetries of the unit cells. 3) The symmetry of the unit cell should reflect the internal symmetry of the crystal.

The unit cell is defined by three basic lattice vectors (a, b and c), with three unit cell dimensions a, b, c and three angles (α, β and γ) between the unit cell vectors.

Symmetry in real space Three types of symmetry operation exist (in addition to translational symmetry): rotation about an axis, mirroring in a plane, and inversion through a center. A rotation axis may be two-fold, three-fold, four-fold or six-fold. Inversion through a center will create a pair of equivalent positions (x, y, z) and (-x, -y, -z). Combinations of these basic symmetry operations will generate further operations. For example, rotoinversion axes (improper rotation), such as -2, -3, -4 and -6 rotations are the combination of rotation about an axis with inversion through a center. Screw axes, such as two-fold (21), three-fold (31 and 32), four-fold (41, 42, and 43), and six-fold (61, 62, 63, 64, and 65) screw axes, arise from the combination of rotation with translation.

Point group “A point group is a group of symmetry operations, all of which leave at least one point unmoved”, as defined by IUCr (65). The compatible combination of non-translational symmetry operations creates 32 crystallographic point groups.

Crystal system and Bravais lattices Crystals can, as we have seen, be classified into 32 point groups. These can further be classified into seven classes, also known as “crystal systems” (Ta- ble 3.1). For example, crystals with only three perpendicular two-fold axes can be described by an orthorhombic unit cell. Taken into account the seven crystal systems and possible translational symmetries, crystals can be divid- ed into 14 Bravais lattices, as given in Table 3.1.

Table 3.1 Crystal systems, characteristic symmetries, and unit cell restrictions Crystal system Bravais Characteristic symmetries Unit cell restrictions type(s) Triclinic P None None Monoclinic P, C Only one 2-fold axis α = γ = 90° Orthorhombic P, I, F, C Only three perpendicular α = β = γ = 90 2-fold axes Tetragonal P, I Only one 4-fold axis a = b, α = β = γ = 90° Trigonal P (R) Only one 3-fold axis a = b, α = β = 90°, γ = 120° Hexagonal P Only one 6-fold axis a = b, α = β = 90°, γ = 120° Cubic P, F, I Four 3-fold axes a = b = c, α = β = γ = 90°

Space group Space group can also be considered as the combination of one of the 32 point groups (without a translational component) with all possible translational components. There are 230 possible combinations, giving 230 space groups, which can be denoted by either short or full Hermann-Mauguin symbols. These notations consist of two parts: (i) a letter indicating the type of Bravais lattice, and (ii) a set of characters or numbers indicating the symmetry elements. The short symbols of symmetry elements are usually used. For example, the full symbol of space group P21/m is P121/m1, considering the b-axis to be unique. The symbol indicates that there is a 21 screw axis along the b-axis and a mirror perpendicular to the b-axis.

3.1.2 Reciprocal space All of the structural information of a crystal is present in reciprocal space. The commonly seen electron diffraction pattern and powder X-ray diffraction data provide structural information in reciprocal space. The structural information from a perfect crystal is concentrated at discrete points that are periodically distributed in reciprocal space. The relationship between real space and reciprocal space is the Fourier transform. Any crystalline material can be presented in reciprocal space with its own unit cell (a*, b*, c*, α*, β* and γ*) and symmetry. The unit cell in reciprocal space is related to the unit cell in real space by Equation 3.1. For example, a* is perpendicular to the bc plane in real space, while the angle between a* and a is zero in cubic, tetragonal and orthorhombic crystal systems. In such systems, a = 1/a*. The unit cell in real space can be deduced from the unit cell in reciprocal space. A lattice point in reciprocal space, ghkl, can be described by Equation 3.2, where h, k and l are integers. a* a  b* b  c* c 1

a* b  a* c  b* a  b* c  c* a  c* b  0 Eq. 3.1

* * * ghkl  ha  kb  lc Eq. 3.2 Diffraction data is collected in reciprocal space. The unit cell of any crystalline material in real space can be deduced from the unit cell observed in reciprocal space. The symmetry in real space of a material can also be obtained by analyzing the data collected in reciprocal space. For example, if the values of h+k for all hkl indices are even, the structure is C-centered. The unit cell parameters, space group and the structure factor amplitude of each reflection can be obtained in reciprocal space. To obtain more detailed structural information, such as the positions of the T and O atoms in a zeolite, it is necessary to obtain a set of experimental diffraction data known as the “structure factors”, denoted F.

3.1.3 Structure factors

The IUCr defines the “structure factor” Fhkl as “a mathematical function that describes the amplitude and phase of a wave diffracted from crystal lattice planes characterized by Miller indices h, k, l (66)”.

Fhkl   f j exp2ihx j  kyj  lz j  j

  f j cos2ihx j  kyj  lz j  i f j sin2ihx j  kyj lz j  j i

 Ahkl iBhkl  Fhkl exp(iahkl ) Eq. 3.3 where the sum is over all j atoms in the unit cell, xj, yj and zj are the fractional coordinates of the jth atom, fj is the scattering factor of the jth atom, and 2 αhkl is the phase of the structure factor. Moreover, │Fhkl│ is proportional to the intensity measured in the experiment. From Equation 3.3, we note that: (1) For a known structure, the amplitudes and phases of the structure factors can be calculated from the Fourier transform of the structure. For an unknown structure, only the amplitudes can be deduced from the diffraction data, and the phases are lost during the diffraction experiment. (2) The structure factor connects structural information in real space and reciprocal space. Every atom in a unit cell in real space will contribute to the intensity of every reflection in reciprocal space. (3) Structure factors can be depicted as vectors in an Argand diagram as shown in Figure 3.1. Summing vector contributions from each atom in a unit cell will give a final Fhkl. It is important to note that the phase of Fhkl is close to the phase obtained from the vector sum of only the heaviest atoms.

(4) If the phases of all the Fhkl are known, the 3D electron density (in the case of X-ray scattering) or the electrostatic potential (in the case of electron diffraction) can be calculated by calculating the inverse Fourier transform according to Equation 3.4: 1 Eq. 3.4 (x, y, z)  Fhkl exp 2i(hx  ky lz) V hkl

Figure 3.1 Structure factor Fhkl represented in an Argand diagram. Figure 3.2a shows an HRTEM image of an inorganic compound Li2NaTa7O19 (space group Pbam, a = 15.23 Å, b = 23.57 Å, c = 3.84 Å) (67) along the c-axis after image processing by CRISP (68). The plane group in this projection is pgg, and Table 3.2 lists the amplitudes and phases of the 28 strong reflections obtained from the HRTEM image. If the phase of the strongest reflection 4 0 0 (highlighted in Table 3.2) is changed from 0° to 180°, the image of this projection changes dramatically (Figure 3.2b). If the phase of the strongest reflection is left unchanged while its amplitude is changed to a third of the original amplitude, the main structural features of this projection are retained (Figure 3.2c). We conclude that the phases of the strong reflections are especially important for the structure determination, and that the phase of a structure factor is much more important than its amplitude. This point will be further emphasized in Chapter 5, where the structures of the RHO zeolite family are determined and predicted.

Figure 3.2 (a) HRTEM image of Li2NaTa7O19 taken along the c-axis after imposing the pgg symmetry. The 28 strongest hk0 reflections obtained from the HRTEM image are listed in Table 3.2. (b) The same projection after changing the phase of the strongest reflection 4 0 0 from 0° to 180°; (c) The same projection after reducing the amplitude of the strongest reflection 4 0 0 from 9641 to 3214 while keeping the phase at 0°.

3.1.4 Structure determination by diffraction Several procedures are available when an unknown structure is to be determined by diffraction. Diffraction data is collected and used to determine the unit cell parameters. The intensities of the reflections are extracted. The space group is then determined from the intensities of the reflections. The next step is to determine the phases of the structure factors, which is a key step in structure determination by diffraction. Several conventional methods are available to determine the phases, such as the Patterson method and various direct methods. A newly developed method known as “charge flipping” is a powerful alternative method for phasing (69–73). Section 3.1.5 presents also a new method for phasing, known as the “strong reflections approach”, based on a known structure. If the phases of the strong reflections are cor- rectly determined, the electron density or electrostatic potential map calculated by the inverse Fourier transformation using these strong reflections will represent the major part of the structure. Chemical information can aid in interpreting the electron density map, and an initial structural model obtained. The final step in the structure determination is to refine the initial structural model against the experimental data. .

Table 3.2 List of amplitudes and phases of 28 strongest reflections extracted from the Fourier transform of the HRTEM image of Li2NaTa7O19 taken along the c-axis, shown in Figure 3.2.

h k l Amplitudes Phases (°) 4 0 0 9641 0 1 2 0 9126 0 0 6 0 8562 0 1 7 0 8118 0 3 5 0 6511 180 3 3 0 5658 180 4 3 0 4698 0 4 4 0 4573 0 1 3 0 4171 180 2 6 0 4096 0 3 2 0 3882 180 2 4 0 3564 0 0 4 0 3518 0 3 4 0 3270 0 2 3 0 2571 180 0 8 0 2511 180 2 1 0 2448 180 0 2 0 2335 0 3 1 0 2330 0 4 6 0 2259 180 1 4 0 2237 180 3 7 0 2207 180 2 0 0 1894 0 5 4 0 1883 0 4 5 0 1850 180 2 2 0 1835 0 5 3 0 1761 180 5 1 0 1666 0

3.1.5 Algorithms for the structure determination The most commonly used four approaches for phasing are introduced in the following section: Patterson methods, direct methods, charge flipping, and the strong reflections approach.

The Patterson method Patterson suggested in 1934 that the following equation could give important information about the crystal structure, without knowing the phases of the structure factors. P(u,v,w) will have peaks that correspond to each interatomic vector in the structure (74).

1 2 P(u,v,w)  F exp 2i(hu  kv lw) Eq. 3.5 V  hkl hkl In conventional direct space, the positions of atoms are defined by the values of the ρ function, which is a function of the fractional coordinates x, y, z in the unit cell. In Patterson space, the vectors between each pair of atoms are defined by generic coordinates u, v and w in the same unit cell. In this way, any pair of atoms located at x1, y1, z1 and x2, y2, z2 will give a peak in Patterson map at generic coordinates u, v and w, where the coordinates are given by:

u = x1 - x2; v = y1 - y2; w = z1 - z2 However, it was not known how to obtain the atomic positions in the crystal from the coordinates of maxima in the Patterson map until David Harker (1906-1991) discovered a method to analyze the Patterson function. Harker discovered that it is not necessary to investigate all the peaks in Pat- terson map: it is sufficient to focus on special locations with high values in the Patterson map. For instance, for a compound crystalized in the space group P21/c, any atom located at (x, y, z) will have three symmetry-related atoms at (-x, -y, -z), (x, 0.5 - y, 0.5 + z) and (-x, 0.5 + y, 0.5 - z) in the unit cell. Vectors between these atoms in Patterson space will be <2x, 2y, 2z>, <0, 2y-0.5, 0.5> and <2x, 0.5, 2z-0.5>. Once the strong peak in Patterson map is identified with the v coordinate being equal to 0.5, for example <0.3, 0.5, 0.1>, the x and z coordinates of heavy atoms can be calculated as follows: 2x = 0.3, 2z - 0.5 = 0.1; x = 0.15, z = 0.3. The y coordinate of the heavy atom can be calculated by the same approach.

Direct methods Hauptman and Karle shared the Nobel Prize in Chemistry in 1985 for their contributions to the development of direct methods for the determination of crystal structures (75). The key step was to develop a practical approach using the Sayre equation.

The Sayre equation was first put forward by David Sayre in 1952 (76). This equation (Equation 3.6) describes how the structure factor of reflection h k l can be calculated as the sum of the products of pairs of structure factors whose indices sum to h k l. For centro-symmetric structures, the phases of the structure factors are restricted to 0° and 180°. The phase relation is described in Equation 3.7. The symbol ≈ is used to denote the fact that there are certain probabilities that the following triplet relationship is true.

Eq. 3.6 Fhkl  Fh'k'l'Fhh', k-k' ,ll' h'k'l'

h,k,l h'k'l' hh', k-k',ll'  0 Eq. 3.7

Fhkl Eq. 3.8 Ehkl  ε F 2 hkl Another important concept, the normalized structure factor (Ehkl) (Equa- tion 3.8), is introduced here, where ε is the enhancement factor and 2 2 <│Fhkl│ > is the average value of │Fhkl│ within a certain resolution shell. As previously stated, strong reflections are particularly important for solving crystal structures, and calculating normalized structure factors is a method used to determine which reflections are strong reflections. Normally, reflections with an E value larger than 1.5 are considered to be strong reflections, and are used in triplet relations for phasing. It is necessary to fix the phases of some strong reflections in order to fix the origin of the unit cell. Most triplets are generated using strong reflections, and the structure factor phases from all possible combinations of these are refined by utilizing what is known as the “tangent formula”. Phased reflections are sorted according to their FOM (figure of merit) value. An electron density map is calculated from the structure factors with phases that have high FOM values. Chemical information can be used to aid the interpretation of this electron density map, locating the positions of the individual atoms in the unit cell.

Charge flipping The charge flipping method uses a dual space iterative phasing algorithm. This algorithm has been applied to single crystal X-ray diffraction (SCXRD) data by Oszlányi and Sütő (69–71) and to powder X-ray diffraction data (PXRD) by Wu (72) and Baerlocher et al (73, 77). Six zeolite framework structures have been solved by this algorithm (78). The procedures are as follows: (1) Random phases are assigned to the experimental amplitudes, generating a random 3D electron density map.

(2) The random 3D electron density map is modified by changing (flipping) the signs of all densities below a user-defined threshold  (a small positive number). (3) The Fourier transform of this modified electron density map is calculated, to give a set of reflections with calculated phases and amplitudes. (4) The calculated amplitudes are replaced by the experimental amplitudes and while the calculated phases are kept for the reflections. (5) A new electron density map is calculated from the new set of reflections and the procedure is repeated. (6) In the case of SCXRD, the iteration is stopped when the calculated amplitudes match the experimental ones. In the case of PXRD, however, the peak overlap problem (Section 3.2.2) makes it necessary to provide an additional histogram. This histogram contains the chemical composition of the material (and ensures that the number and heights of the peaks in the map correspond to the chemical formula).

The strong reflections approach The Patterson method and direct methods attempt to solve the phase problem in reciprocal space, while the charge flipping method alternates between reciprocal space and direct space (it is a “dual-space” method). The strong reflections approach, initially developed for the structure determination of quasi-crystal approximants by Hovmöller and Zou’s group, “borrows” phases from a related known structure and uses these to solve the unknown structure (79). Related known structures are identified based on them having similar distributions of strong reflections (which carry important structural information). The empirical structure solution of a series of quasicrystal ap- proximates has shown that, if the distributions of strong reflections are similar, the corresponding phases are similar. For example, the structure of 2 τ -Al13Co4 was solved based on the known structure of the related m-Al13Co4 (79). The steps were as follows:

(1) Strong reflections were selected from the known m-Al13Co4 structure. 2 (2) New indices for reflections of the unknown τ -Al13Co4 were obtained by scaling indices of reflections from known m-Al13Co4 according to the relationship of their unit cells. 2 (3) A 3D electron density map of τ -Al13Co4 was calculated using amplitudes and phases of structure factors from the known m-Al13Co4 structure. (Section 3.1.3 showed that a structure can be solved even if the amplitudes are inaccurate, so not only the phases were borrowed from the known structure, but also the amplitudes.) (4) Chemical knowledge was used to identify atoms in the resulting electron density map. The application of the strong reflections approach to other systems will be described in Chapter 5.

3.2 Structure determination of zeolites

3.2.1 Single crystal X-ray diffraction (SCXRD) An unknown form of radiation that was able to penetrate opaque bodies was discovered by the German scientist Wilhelm Conrad Röntgen (1845-1923) at the University of Würzburg in 1895. The radiation was named “X-rays”, and Röntgen was awarded the Nobel Prize in Physics in 1901 for this discovery (80). An application of X-rays in the fields of physics, chemistry and biology came from Professor Max von Laue (1879-1960), who won the Nobel Prize in Physics 1914 for his discovery of the diffraction of X-rays by crystals (81). Benefiting from the pioneers’ discoveries, William H. Bragg (1862-1942) and William L. Bragg (1890-1971) (father and son) together won the Nobel Prize in Physics in 1915 for their services in the analysis of crystal structures by means of X-rays (82). The SCXRD technique has developed very rapidly, and is now a mature technique for structure determination. The technique can be applied on microcrystals (with dimensions of several micrometers) when synchrotron light sources are used. The basic procedures of structure determination by this method have been described in Section 3.1.4. About half of all zeolite structures have been solved by SCXRD (78).

3.2.2 Powder X-ray diffraction (PXRD) SCXRD cannot be used if only nano- and submicrometer-sized crystals are available. Silicate-based zeolites have wide industrial applications but they often form polycrystalline materials too small to be studied by SCXRD. PXRD can be a powerful technique in such cases. This technique is widely used for phase identification and has been used also for structural analysis of these polycrystalline materials. 116 zeolite structures have been solved from PXRD data (78). The steps for structure determination by PXRD are similar to those described in Section 3.1.4. However, the presence of impurities makes the unit cell determination much more challenging. Additional information from, for example, the ICDD database of experimental powder diffraction patterns (http://icdd.com/products/) or scanning electron microscopy (SEM) images of the samples can be used to identify impurities and aid in the determination of the unit cell (78). Structure determination by PXRD is more difficult than it is using SCXRD not only because of the phase problem, but also because of the no- torious ambiguity in the assignment of intensities caused by the overlap of reflections. In the cubic system, for example, the d values of the (5 5 0), (5 4 3) and (7 1 0) reflections are identical. Thus, these three peaks in the PXRD pattern coincide, and only the sum of the intensities of these three reflections

can be measured. Since strong reflections are important for phasing, structure determination by PXRD is hindered by peak overlap. Direct methods, developed originally for SCXRD, can be applied to PXRD data, if peak overlap is not severe. The first zeolite analogue structure solved by direct methods from PXRD data was an aluminophosphate molecular sieve, AlPO-12 (FTC: ATT) in 1986 (83). Two years later, Lynne B. McCusker solved a zeolite structure (Sigma-2, FTC: SGT) from PXRD by direct methods (84). Two further breakthroughs in structure determination by PXRD are the aforementioned charge-flipping algorithms and the FO- CUS program, which is designed to solve the structures of zeolites using PXRD data (78, 85, 86). However, structure solution of zeolite crystals with high levels of disorder or poor crystallinity using PXRD data is much more difficult.

3.2.3 FOCUS The FOCUS computer program (http://www.iza-structure.org/ under Other Links: Software) was initially written to determine the structures of zeolites from PXRD data. After data collection, peak search, indexing, and space group determination, the amplitudes of the structure factors are extracted from the PXRD data and given random phases. A 3D electron density map is generated from these structure factors. The program uses chemical information about zeolite bond lengths and angles to search for the largest frag- ments of zeolites via a backtracking algorithm, and creates a new set of amplitudes and phases. It then combines the new phases with the experimental amplitudes, and calculates a further 3D electron density map. This cycling procedure is repeated until the phases converge. Any solution that resembles a zeolite is registered in a histogram after each cycle, and a new set of random phases will then be assigned to the experimental amplitudes to search for new solutions in a next cycle. A dominating solution will often appear from different sets of random starting phases. This is then considered to be the most probable candidate for the structure of the unknown zeolite (78). Figure 3.3 shows a flow chart of how the FOCUS program operates. Recent developments in the program have allowed electron diffraction data to be used to determine structures in a similar manner (87). One classical example of structure determination from electron diffraction data by FOCUS is demonstrated on SSZ-87 (88). It is worth mentioning that in this case only 15% of the complete electron diffraction data was collected, which was enough for the structure solution. 22 zeolite frameworks have been solved by FOCUS (78).

Figure 3.3 Flow chart of the FOCUS program. By courtesy of Dr. Stef Smeets.

3.2.4 Rotation electron diffraction (RED) Our research group has focused on the method development of electron crystallography, including electron diffraction and HRTEM, for many years. Electron crystallography does not suffer from the problems that face X-ray crystallographers, such as crystals that are too small to analyze, the presence of impurities, peak overlap, and a high level of disorder in the crystals. This section describes one breakthrough in electron crystallography, the development of the rotation electron diffraction (RED) technique in our research group. The term “RED” is not only an abbreviation of “rotation electron diffraction”, but also the name of a computer program package for the automated collection and processing of 3D electron diffraction data (31, 32). The collection of electron diffraction data is controlled by the RED-data collection program, and is similar to the data collection procedure used in SCXRD. Data can be collected from tiny crystals (of dimensions as small as 100 nm) by electron diffraction. 3D electron diffraction data are collected at small angular intervals over a large range of angles, which is achieved by combining fine tilts (0.05°-0.20°) of the electron beam with coarse tilts (2.0°-3.0°) of the goniometer. The RED-data collection program can be installed on conventional TEMs without any hardware modifications. Data collection by RED can start from an arbitrary orientation of the crystal, and approximately

480 frames can be collected within half an hour (using a goniometer tilt increment of 2.0° and an electron beam tilt increment of 0.20°). Experimental electron diffraction data can be reconstructed into a 3D dataset by the RED-data processing program. The ED frames that have been collected are automatically merged into a 3D dataset under the control of the user, who can define a set of parameters. The unit cell of the crystal is determined from the positions of the diffraction spots in the ED frames. The list of reflections with their indices and intensities are output to an HKL file in a standard HKLF4 format defined in the SHELX program suite (89). The RED-data processing program contains also statistical analyses that assist the user in determining the space group of the crystal. A unit cell refinement function has recently been added to the program, which can determine the unit cell parameters to accuracies of 0.01 Å and 0.1°. Structure determination methods that have been used for X-ray diffraction, including programs that use direct methods (programs SHELX (89) and SIR (90)), charge flipping (programs SUPERFLIP (91), Jana (92)), simulated annealing (programs FOX (93) and SIR (90)) and FOCUS (FOCUS (85, 86), especially for zeolites), can be applied directly on RED data. A variety of structures including zeolites, zeolite-related materials, metal- organic frameworks (MOFs), covalent organic frameworks (COFs), germi- nates and quasi-crystal approximants have been solved from RED data. Many zeolite structures have been solved from RED data. They are ITQ-51 (FTC: IFO) (34), SSZ-61 (*-SSO) (94), SSZ-87 (IFW) (88), SSZ-45 (EEI) (95), EMM-23 (*-EWT) (96), PKU-16 (POS) (97), ITQ-53 (38), ITQ-54 (- IFU) (33), CIT-7 (CSV) (98), ZSM-25 (MWF) (99) and PST-20 (99).

3.2.5 HRTEM We have seen that the phase information of structure factors, which is essen- tial for structure determination, is lost in diffraction methods such as SCXRD, PXRD and RED. They are, however, preserved in HRTEM images. The structure can be determined from a 3D reconstruction of HRTEM images of different projections (100). Another advantage of HRTEM is that it is possible to analyze crystal structures with defects. Intergrowth and stacking disorders are often observed in zeolites, which makes it difficult to determine the structure. No standard methods for analyzing this kind of disordered material are currently available. The origins of the disorder and the distribution of ordered domains can, however, be observed in HRTEM images (101). Zeolite beta was first synthesized by the researchers at Mobil Oil Corpo- ration. This zeolite has intergrowth of two polymorphs, A and B, and the structure was determined by HRTEM, electron diffraction and computer- assisted modelling (102, 103). The unit cell and possible space groups were determined from a series of selected area electron diffraction (SAED) pat-

terns. The high quality HRTEM image along the [100] direction shown in Figure 3.4 reveals the 4-, 5-, 6- and 12-rings along this projection (101). This information allowed a structure model of Zeolite beta to be proposed. Two pore stackings occur, ABABAB and ABCABC, which gave visible effects in the diffraction pattern. The SAED pattern taken along [100] has diffuse streaks parallel to the c*-axis for reflections with h ≠ 3n or k ≠ 3n, and sharp spots for reflections with h = 3n or k = 3n. These effects arise from the stacking disorder of the layers (ab planes) along the c*-axis with successive layers shifted one-third of the a or b parameter relative to the previous layer (101).

Figure 3.4 HRTEM image of Zeolite beta taken along the [100] direction and showing 12-ring channels. The insert shows the corresponding SAED pattern with streaks perpendicular to the building layers, parallel to the stacking direction c*. By Courtesy of Dr. Jie Su at Stockholm University.

3.2.6 Model building Model building was widely used to explore new zeolite structures in the early days, and is still useful when other methods fail. The structure of ITQ- 3 (FTC: ITE) (104), for example, was solved by Avelino Corma’s and Paul A. Wright’s groups, who derived its orthorhombic unit cell (a = 20.622 Å, b = 9.724 Å, c = 19.623 Å) from PXRD data. An important part of the solution was the realisation that the building layers in RUB-13 (FTC: RTH) (105) (with a 2D unit cell of dimensions a = 20.530Å, b = 9.762 Å, γ = 90°) were identical to the building layers in ITQ-3. The c dimension of the unit cell of ITQ-3 is twice that of RUB-13, and the structural model of ITQ-3 could be built based on the RUB-13 structure. The building layers in ITQ-3 are related by mirror symmetry, while the building layers in RUB-13 are related through an inversion center. The structural model was confirmed by Rietveld refinement against synchrotron PXRD data. Model building is very time-consuming and its success depends heavily on the experience, patience and ingenuity of the researchers. Chapter 6 summarizes some common structural features of zeolites. We hope that our study will make it easier to solve structures by model building, and that the principles of model building might be implemented in software for automated structure solution.

3.3 Rietveld refinement An approximate structural model that has been obtained by one or several of the methods described above must be refined. This is the process that mini- mizes the difference between diffraction data calculated from the structural model and the experimental data. All the structural models determined in the work described in this thesis have been obtained from RED data or the strong reflections approach and have been subsequently refined against PXRD data. RED data suffers from dynamical effects and electron beam damage, which means that it can be used to solve structures, but is less suit- ed for refinement. Hugo Rietveld proposed a new refinement method in 1969 that does not use the integrated powder intensity, but uses the profiled intensities of the powder diffraction pattern directly from the step-scanning measurements (106). This refinement approach is known as “Rietveld refinement”. It min- imizes a parameter, △, by least-squares, where Δ is obtained from Equation 3.9. obs calc 2 Eq. 3.9   wi Yi Yi  i

obs calc where Yi and Yi are the measured and calculated intensities of the individual profile points in the 2θ scan, respectively, and wi is a statistical weight. calc The profiled Yi intensities are calculated. The angle 2θhkl, which is the angle of the center of the peak, is calculated from h k l indices and the unit cell dimensions. In the ideal case, the diffraction lines are sharp, but experimental factors give them a shape that is broader than the ideal case. This introduces a further parameter that describes the shape of the peak into the calculation. Several functions are available to describe peak shapes, such as Gaussian, Lorentzian, Pearson VII, and pseudo-Voigt. Equation 3.10 shows calc how the Yi are calculated when a Gaussian function is used

calc 2 Eq. 3.10 Yi  2 In2/ Ihkl exp 4In2{( xi - 2hkl )/H} / H hkl

th where xi is the 2θ value of the i profile point; 2θhkl is the angle of the center of the peak; H is the full-width at half-maximum height (FWHM) of the peak; Ihkl is the intensity of the reflection with the indices h k l. Fhkl can be calculated by taking the Fourier Transform of the structural 2 model. Fhkl is proportional to Ihkl, taking a number of additional experimental effects into account. These include the multiplicity (j), the polariza- tion factor (P), Lorentz factor (L) and absorption (A). Equation 3.11 shows how Ihkl can be calculated.

2 Ihkl  cjPLAFhkl Eq. 3.11 where c can be other factors, such as temperature, sensitivity and detector etc. The success of Rietveld refinement can be determined in several ways. It is important, of course, that the structure after the refinement should make chemical sense, with reasonable values for parameters such as bond angles, bond lengths and hydrogen bonding profile. Further, the profile residual Rp (Equation 3.12) and weighted profile residual Rwp (Equation 3.13) should converge, and should normally both be less than 10%.

Y obs Y calc  i i Eq. 3.12 R  i p obs  Yi i 2 0.5  w Y obs  Y calc    i i i  Eq. 3.13 R  i wp  obs 2    wi Yi    i 

Rexp (Equation 3.14) gives an indication of the best possible agreement value for Rwp that can be obtained:

0.5    n  p  Eq. 3.14 R  exp  obs 2   wi Yi    i  where n is the number of observations and p is the number of parameters refined in the least-squares procedure.

obs calc 2 w Y Y 2  i  i i   R  2 i wp GOF       Eq. 3.15 n  p Rexp 

The goodness of fit (GOF), defined as the ratio between Rwp and Rexp, is another measure of the success of refinement (Equation 3.15). The value of GOF is unity for a perfect model (which is, of course, impossible to achieve in practice). The structural models described in this thesis have been refined against PXRD data. Initial structural models have been obtained from RED data or by the strong reflections approach, and refined by Rietveld refinement. We have thus shown that RED data and the strong reflections approach can be used to solve crystal structures. More detailed structural information, such as the locations of guest molecules and interactions between the framework and non-framework species, was obtained from the models after Rietveld refinement.

4. Structure determination of zeolites and zeolite-related materials by rotation electron diffraction (RED)

Zeolites and zeolite-related materials tend to crystallize as polycrystalline powders, which makes it difficult to determine the structures of these materials by conventional SCXRD. In addition, peak overlap, poor crystallinity and impurities make it difficult to determine their structures by PXRD. Elec- tron crystallography, which includes electron diffraction and HRTEM, can overcome these difficulties. Our research group has developed a new method, known as “rotation electron diffraction” (RED), in which electron diffraction data are collected automatically and used to solve structures. Four structures solved from RED data are presented in this chapter.

4.1 COE-3 and COE-4 (Paper I) COE-3 and COE-4 (where the abbreviation “COE” is used to denote that the materials were first prepared at the International Network of Centers of Ex- cellence) are post-synthesis materials that are prepared from layered silicate RUB-36 and dichlorodimethylsilane. The structure of RUB-36 with FER layers (107) is connected after dichlorodimethylsilane is introduced, generating a new 3D framework (108). This as-made sample is denoted “COE-3”, while its calcined form is denoted “COE-4”. The two methyl groups in the COE-3 structure become two hydroxyl groups after calcination. Pure RUB- 36 generates a framework of CDO type after the condensation, and thus COE-3 and COE-4 belong to a family of interlayer-expanded zeolites (IEZ) with CDO structures. The structures of COE-3 and COE-4 were initially determined by model building, and the structural models were refined against PXRD data. Mean- while, Prof. Osamu Terasaki contributed to the structure determinations of these IEZ structures using HRTEM images (109, 110). However, due to electron beam damage and the fact that the crystals take up only preferred orientations on the TEM grids, it was extremely difficult to collect high quality HRTEM images. Hence, we decided to use the RED technique to solve

the structures of these plate-like materials that are sensitive to the electron beam.

Figure 4.1 (a-c) 3D reciprocal lattices of COE-3 reconstructed from three RED datasets. Crystals from which the RED data were collected are shown as inserts. (d-l) The 2D slices (h0l), (hk0) and (0kl) cut from Dataset 1COE-3, Dataset 2COE-3 and Dataset 3COE-3. Reproduced from Ref. (111) with permission from the Royal Society of Chemistry. Copyright © 2014, Royal Society of Chemistry. Three RED datasets were collected from three different crystals of COE- 3. The electron diffraction (ED) frames collected over a range of tilt angles were processed for each dataset into a 3D lattice in reciprocal space (Figure 4.1a-c). The three datasets were complementary to each other. Two-

dimensional slices, (h0l), (hk0) and (0kl), were cut from the Datasets 1COE-3, 2COE-3 and 3COE-3 (Figure 4.1d-f, g-i, and j-l, respectively). The reflection conditions showed that the possible space groups of COE-3 are Cmc21 (No. 36), C2cm (No. 40) and Cmcm (No. 63). Most of the zeolite structures in the IZA Database of Zeolite Structures are centrosymmetric, so we took as a working hypothesis that Cmcm was the correct space group, and used it in the structure determination (111). The unit cell parameters were from Da- taset 1 (a = 7.2 Å, b = 21.8 Å, c = 13.6 Å). Unfortunately, the structure of COE-3 could not be solved from any of the three RED datasets alone. We decided, therefore, to merge the datasets to increase the data completeness. It was now possible to solve the structure of COE-3 by direct methods, using the SHELX program. All the five Si atoms, including the bridging Si atoms (Si5) between the FER layers, and five of eight O atoms in the asymmetric unit could be identified in the electron density map. The missing O atoms were added manually according to the SiO4 tetrahedral geometry. Solid-state NMR had shown that the bridging Si5 is tetrahedrally coordinated and is bound to another two methyl groups. These methyl groups were added manually, and the structural model was refined against the RED and PXRD data. The refinement against the RED data converged to an R1 value of 0.38 for the 227 reflections. Figure 4.2a shows Rietveld refinement plots of COE-3. The final Rietveld refinement against PXRD data converged to reasonable R values (Rp: 0.033, Rwp: 0.043 and GOF: 2.475). Two RED datasets were collected from COE-4 (Figure 4.3). Again, it was not possible to solve the structure of COE-4 from a single dataset. The two datasets were thus merged, which enabled us to obtain a structural model of COE-4. All the five Si atoms and six out of nine O atoms in the asymmetric unit were identified in the electron density map. The missing O atoms were added manually according to tetrahedral geometry. Rietveld refinement against PXRD data converged to reasonable R values (Rp: 0.038, Rwp: 0.048 and GOF: 1.988). Figure 4.2b shows Rietveld refinement plots of COE-4. After the insertion of the linker groups between the FER layers, the pore opening of the channels in the COE-3 and COE-4 increased from 8-ring (in the CDO type framework) to 10-ring along a- and c-axes. Due to the mirror symmetry perpendicular to the a-axis, the bridging Si5 atoms in COE-3 and COE-4 are disordered with the maximum occupancy of 0.5, as shown in Figure 4.4.

Figure 4.2 Observed (blue), calculated (red) and difference (black) PXRD profiles for the Rietveld refinement of the COE-3 (a) and COE-4 (b) structures. The vertical bars indicate the positions of Bragg peaks (λ = 1.5418 Å). Reproduced from Ref. (111) with permission from the Royal Society of Chemistry. Copyright © 2014, Royal Society of Chemistry.

Figure 4.3 (a-b) 3D reciprocal lattices of COE-4 reconstructed from two RED datasets. The reciprocal lattice axes are marked with a*, b* and c* in red, green and blue, respectively. The crystals from which the data was collected are shown as inserts. (c) 2D slice of the (0kl) plane cut from Dataset 2COE-4. (d-f) (h0l), (hk0) and (0kl) planes cut from Dataset 1COE-4 (a). Repro- duced from Ref. (111) with permission from the Royal Society of Chemistry. Copyright © 2014, Royal Society of Chemistry.

Figure 4.4 Projections of COE-3 (a) and COE-4 (b) along the c-axis. The disordered Si atoms are highlighted inside green circles.

4.2 EMM-9 (Paper II) A fluoroaluminophosphate, EMM-9, synthesized using 4- (dimethylamino)pyridine as an OSDA in the fluorine medium, was reported in 2010 (112). Its structure, however, remained unsolved for more than six years. It was difficult to solve the structure of EMM-9 because convincing unit cell parameters could not be determined from the PXRD data. We decided, therefore, to collect a RED dataset from EMM-9, and use it to solve the structure. The unit cell dimensions (a = 6.98 Å, b = 13.57 Å, c = 14.76 Å, α = 90.6°, β = 103.0°, γ = 89.5°) were determined from the reconstructed 3D reciprocal lattice (Figure 4.5a) using strong basic reflections. The unit cell parameters indicated that the EMM-9 structure is monoclinic. Weak reflections are present along a* in the reciprocal lattice, shown in yellow in Figure 4.5a. These weak reflections were ignored during the initial structural investigation. The systematic absence was deduced as 0k0: k = 2n, from three 2D slices (hk0), (0kl) and (h0l) (Figure 4.5b-d). Possible space groups are P21 (No. 4) and P21/m (No. 11). The centrosymmetric space group P21/m was taken as a working hypothesis, and used in the structure determination. A structural model of EMM-9, based on an average tetrahedral (T) atom (where T is Al or P) and O, was constructed using phases obtained by direct methods using the SHELX program. All of the T atoms and eight out of ten O atoms in the asymmetric unit could be identified in the electron density map. The weak super-lattice reflections were attributed to the alternation of AlO4 and PO4 units, which led to a doubled length of the a-axis and the space group of P21/m rather than P21/a.

Figure 4.5 (a) The 3D RED reciprocal lattice seen along the b*-axis. Super- lattice reflections are shown in yellow. Without considering the super-lattice reflections, the unit cell parameters are a = 6.98 Å, b = 13.57 Å, c = 14.76 Å, α = 90.6°, β = 103.0°, γ = 89.5°. (b-d) Three 2D slices (h0l) (b), (hk0) (c) and (0kl) (d) cut from the 3D reciprocal lattice. The structural model was further refined against synchrotron data, with the super-lattice reflections included. Rietveld refinement of EMM-9 was used to complete the structure, locating the OSDAs whose locations were still unknown. The models were adjusted to ensure that they were chemically realistic and to obtain reasonable R values. The procedures used were as follows: 1) The structural model was determined and high-angle X-ray diffraction data were used to find the scale factor. We decided that if the high-angle data could be scaled accurately, the scale factor determined should be used for the complete diffraction pattern.

2) No other parameters, such as peak shape and unit cell parameters were refined. We ran the software once only, and revealed that the calculated and observed PXRD data differed greatly (Figure 4.6a). The difference Fourier map was calculated (Figure 4.6b), and showed a cloud of electron density between layers. The individual atoms of the OSDA could not be resolved, but we were able to obtain approximate positions. 3) The software was run once again, now including both the layers and the positions of the OSDAs. The differences between the calculated and observed PXRD data were smaller (Figure 4.6c), and the positions of F ions could be determined from the model (Figure 4.6d). 4) The differences between calculated and observed PXRD data were further reduced by refining the atomic positions, unit cell parameters, peak shape function etc. The final refinement of EMM-9 converged to reasonable R values (Rwp: 0.104, Rp: 0.079 and GOF: 1.333). Figure 4.7 shows the final Rietveld refinement plots.

Figure 4.6 (a-b) Powder X-ray diffraction traces (a) and difference electron density map (b) to find positions of OSDAs through Rietveld refinement. (c- d) Powder X-ray diffraction traces (c) and structural model (d) used to locate positions of F ions. (λ = 0.866 Å).

Figure 4.7 X-ray Rietveld refinement plot of EMM-9. The observed, calculated and difference curves are shown in blue, red and black, respectively. The vertical bars indicate the positions of Bragg peaks (λ = 0.866 Å). EMM-9 is a layered fluoroaluminophosphate structure. The asymmetric unit comprises four Al atoms, four P atoms, two OSDAs, two fluorine ions and one guest water molecule. The CBUs in the EMM-9 structure are closely related to sti (Figure 4.8 a-b). The sti CBU, described as double 4-rings with one edge disconnected (Figure 4.8c), has been introduced in Section 2.1.1 as the 4-4- SBU. These CBUs are connected through O7 and O14, generating an sti chain along the a-axis (Figure 4.8d). The sti chains are further assembled to a layer through O1 and O8 (Figure 4.8e). This layer is denoted as “SFO”, since it has been identified in the 3D SFO framework type. Organic templates interact through π-π bonds (Figure 4.9a) and the OSDAs interact with the layers through hydrogen bonds (Figure 4.9b). Translation of adjacent layers by 1/2b and 1/3a followed by condensation along the c-axis (Fig- ure 4.10 a-d) produces a 3D SFO framework type (Figure 4.10 e-f).

Figure 4.8 Stick (a), polyhedral (b) and simplified (c) models of sti CBU in the EMM-9 structure. The sti CBUs are connected by O7 and O14 along the a-axis (d) and by O8 and O1 along the b-axis (e).

Figure 4.9 (a) π-π interactions between organic templates. (b) Hydrogen bonds between organic templates and the inorganic layer.

Figure 4.10 (a) The EMM-9 structure viewed along the c-axis. (b-d) The adjacent layer translated by 1/2b and then -1/3a. (e-f) The formation of the layered EMM-9 to a 3D SFO framework after the condensation along the c- axis.

4.3 EMM-26 (Paper III) EMM-26 is a new borosilicate zeolite with pores of medium size. It has been synthesized using a linear organic dicationic OSDA, and forms crystals of nanometer size. It was difficult to solve the structure from in-house PXRD data due to difficulties in determining the space group. EMM-26 is more stable than both COE-3 and COE-4. Only one RED dataset was needed for the structure determination (Figure 4.11). The unit cell parameters (a = 19.4 Å, b = 15.8 Å, c = 17.9 Å, α = 90.5°, β = 89.7°, γ = 89.8°) and reflection conditions of EMM-26 could be easily deduced from this dataset. The space group with highest possible symmetry is Cmce. The structure was successfully solved from the RED data by direct methods using the program SIR2011 (90). All the seven T atoms and 13 O atoms in the asymmetric unit

were identified in the map. Five of the T atoms (Si1, Si3, Si5, Si6 and Si7) are at general positions, while the rest are at the 8d (Si2) and 8e (Si4) sites. The structural model was refined against the RED data. The final refinement showed three boron-rich positions (Si4: 13%, Si6: 5% and Si7: 37%). The structural model obtained from RED data was further refined against synchrotron PXRD data collected at the ESRF. Rietveld refinement converged to reasonable R values (Rp: 0.051, Rwp: 0.066 and GOF: 3.636). Figure 4.12 shows the final refinement plots. The Rietveld refinement shows that Si4, Si6 and Si7 are also boron-rich positions. The occupancies of boron at these three positions are 14%, 8% and 35%, respectively. These results are consistent with those obtained from the RED data. The position of the OSDA was determined during the Rietveld refinement (Figure 4.13). This is the first time that the preferred positions of boron in a zeolite structure have been determined from RED data.

Figure 4.11 (a) 3D reciprocal lattice of EMM-26 reconstructed from RED data, with the corresponding TEM image of the crystal inserted. (b-d) Three 2D slices (0kl), (h0l) and (hk0) extracted from this dataset. EMM-26 is a medium pore borosilicate zeolite with 2D 10  10 ring channels. The EMM-26 structure contains fer CBUs (Figure 4.14a), which are connected in an up-down configuration along the b-axis to form a fer chain. Figure 4.14b illustrates how adjacent fer chains are further linked by

one extra SiO4 motif to generate a dense layer (Figure 4.14c). The 2D dense layers are related through a mirror plane and connected to form a 3D zeolite framework (Figure 4.14d). The structure contains 2D intersecting elliptical 10-ring channels that run parallel to the dense layers. The HRTEM image (corrected for the contrast transfer function (CTF) using the QFocus program (113)) along the b-axis was also taken, and proved to be consistent with the structural model (Figure 4.15).

Figure 4.12 Rietveld refinement plots of EMM-26. The observed, calculated and difference curves are in blue, red and black, respectively. The vertical bars indicate the positions of Bragg peaks (λ = 0.4009 Å).

Figure 4.13 Location of the OSDA in the cavity. Bridging O atoms have been omitted, for clarity. Boron atoms are distributed among the Si4, Si6 and Si7 positions. Of these positions, Si7 is particularly boron-rich.

Figure 4.14 (a) 1D chain composed of fer CBU. (b-c) The 2D dense layer. (d) The 3D framework created by connecting two dense layers through the mirror symmetry. The bridging O atoms have been omitted for clarity. Boron distributes in the Si4, Si6 and Si7, respectively. Among them, Si7 is a boron- rich position.

Figure 4.15 (a) Selected area diffraction pattern along the [010] direction, and a low magnification image of the crystal shown as an insert. (b) High resolution transmission electron microscopy image along the [010] direction corrected for the contrast transfer function (CTF) using the QFocus program. Inserts are the Fourier transform of the whole image, the averaged image with pmm symmetry (inside the blue square), and the structural model viewed along the b-axis.

4.4 Conclusions The structures of COE-3 and COE-4 have been solved from merged RED data. This technique provides an alternative approach to solving the structures of zeolite-like materials with plate-like morphology and high electron- beam sensitivity. It was easy to determine the structures of the layered fluoroaluminophosphate EMM-9 and the borosilicate EMM-26 using the RED technique. Importantly, this is the first time that it has been possible to locate boron-rich positions in a borosilicate material from RED data.

5. Unravelling the structural coding of the RHO zeolite family

Chapter 4 describes the successful and straightforward structure determinations of COE-3, COE-4, EMM-9 and EMM-26 from RED data. It is necessary to collect experimental RED data with a resolution better than 1.2 Å in order to determine the structures of zeolites and zeolite-like materials. These materials, however, are highly sensitive to the electron beam, and it is not always possible to collect high resolution electron diffraction data from sub- micron-sized zeolite crystals over a large tilting range. New methods are required to solve complicated zeolite structures from RED data with low resolution. This chapter will describe the application of the strong reflections approach, which has been successfully adopted in the structure determination of quasi-crystal approximants (Section 3.1.5), in the zeolite field. This method can unify the identification of a certain zeolite family, the determination of its structure, and the prediction of structures with the similar structural features. This approach can be used, even if the data is of a resolution lower than 1.2 Å. ZSM-25 is used as an example to illustrate the power of this method. This material was synthesized by the researchers at Mobil Oil Corporation more than 30 years ago.

5.1 Structure determination of ZSM-25 (Paper IV) ZSM-25, first reported in 1981, was synthesized using Na+ and tetraethylammonium (TEA+) ions as OSDAs (114). Its Si/Al ratio is around 3.4. Professor Suk Bong Hong (at POSTECH, Pohang, South Korea) has studied on ZSM-25 for more than 15 years (21, 115), and his group has devoted a great deal of effort in the preliminary characterization of this zeolite. ZSM- 25 has characteristic features of small pore zeolites, with a high uptake of CO2 and low uptake of N2 at 295 K. Moreover, the IR and Raman spectra suggest that a large number of 4-rings are present in the structure of ZSM-25. However, its detailed structural information such as unit cell parameters, space group and atomic positions was still unknown when the project described here started. RED data from ZSM-25 was collected in order to explore its structure (Figure 5.1a-c). The unit cell parameters were deduced from this dataset, and they suggested that the crystal was cubic, with a = 42

Å. The highest-symmetry space group compatible with the data is Im-3m. More accurate unit cell parameters were obtained by refining the structure against PXRD data, and the a value was determined to be 45.07 Å, based on the preliminary identification of the space group as Im-3m (Figure 5.2). Un- fortunately, the resolution of the collected RED data is around 2.3 Å, which is too low to determine the structure of this complex compound using conventional phasing methods.

Figure 5.1 RED data of ZSM-25 (top) and PST-20 (bottom). (a), (d) The 3D reciprocal lattices of ZSM-25 and PST-20 respectively, reconstructed from the RED data. Inserts are the TEM images of the crystals from which the RED data were collected. The (h k -h-k) plane of ZSM-25 (b) and PST-20 (e). The (h k h) plane (c) and (h k -k) (f) plane cut from (a) and (d). The distributions of the strong reflections for ZSM-25 and PST-20 are similar. Re- printed with permission from Ref. (99). Copyright © 2015, Rights Managed by Nature Publishing Group.

Figure 5.2 The profile fitting of ZSM-25 (with an X-ray wavelength of 0.63248 Å) Three frameworks, KFI, RHO and PAU, with the same space group were identified in the Database of Zeolite Structures (116). The strong reflections of ZSM-25 are distributed in the same regions of reciprocal space as those calculated for RHO and PAU, indicating that RHO, PAU and ZSM-25 are structurally related, see Figure 5.3a-d and Figure 5.4. Researchers who synthesize zeolites face the problem of knowing whether the substance they have synthesized is a known or unknown zeolite. If the experimental PXRD pattern of a synthesized material fully matches the simulated PXRD pattern of a known zeolite structure, they can conclude that the synthesized material is a known zeolite. However, they have in this case compared the characteristic diffraction information in reciprocal space from the two materials, and the conclusion is based on the assumption that if this information is identical for the two materials, and if the chemical elements present are the same, their structures are the same. If the characteristic diffraction patterns of two structures are similar, showing, for example, a similar distribution of strong reflections, the two structures are related in real space. Moreover, the strong reflections are related to the main structural features of a crystal. We saw in Chapter 3 that the phase information of diffracted beams is not immediately available in a recorded diffraction pattern, but it can be determined by considering the strong reflections. An attempt to phase the strong reflections of ZSM-25 from the known PAU structure was carried out. Twenty-one symmetry-independent strong reflections from the RED data of ZSM-25 were used for the structure solution. The structure factor amplitudes of these 21 strong reflections were extracted from the experimental RED data, and their phases were adopted from corresponding

reflections in the PAU structure (Table 5.1). The 3D electron density map was calculated, and all of the 16 T atoms in the asymmetric unit could be located (Figure 5.3e). The missing O atoms were added manually to complete the ZSM-25 structure (Figure 5.3f). The cubic unit cell contains 1440 T atoms and 2880 O atoms. The structural model has been confirmed by Rietveld refinement against synchrotron PXRD data collected at ESRF (Fig- ure 5.5a). The R values are reasonable, which shows that our structural model is correct.

Figure 5.3 (a-b) The 2D slice of (h k 0) (a) and (h k -h-k) (b) cut from the 3D reciprocal lattice reconstructed from the RED data. The symmetry has been superimposed to allow for a better comparison. (c-d) Simulated diffraction pattern of the idealized PAU structure, with the structure factor phases marked in blue (180°) or red (0°). (e) 3D map generated using amplitudes obtained from the RED data of ZSM-25 and phases calculated from the structure of PAU. (f) The framework structure of ZSM-25. Reprinted with permission from Ref. (99). Copyright © 2015, Rights Managed by Nature Publishing Group. The PAU structure will be described here, to aid in understanding how the ZSM-25 structure was solved. The PAU structure is composed of seven different cages; [4126886] (lta), [4882] (d8r), [41286] (pau), [466286] (t-plg), [4583] (t-oto), [4684] (t-gsm) and [4785] (t-phi) (Figure 5.5b). Three of them (lta, d8r and pau) build the scaffolds along the unit cell edge (Figure 5.5c). The maximum ring size for all the cages is 8, which shows that PAU is a

small pore zeolite. The body-centering of the PAU structure means that the structure contains two such interpenetrated scaffolds. The lengths of the lta, d8r and pau cages are approximately 12 Å, 3 Å and 7 Å, respectively. The unit cell parameter of the PAU structure is approximately 35 Å (35 Å = 12 Å +3 Å + 10 × 2 Å). The other four cages are embedded in the free space between these two scaffolds (Figure 5.5e). The scaffolds of the ZSM-25 structure can be constructed by inserting a pair of d8r and pau cages (which has a dimension of approximately 10 Å) into the scaffolds of the PAU structure. This makes the connection sequence between the two lta cages lta-d8r-pau-d8r-pau-8dr-pau-8dr-lta (Figure 5.5d). The unit cell increases from 35 Å in PAU to 45 Å in ZSM-25. The empty space between the two scaffolds is filled by t-plg, t-oto, t-gsm and t- phi cages (Figure 5.5f). The common structural features between PAU and ZSM-25 are that (1) they possess the same building units, (2) the two expanded interpenetrated scaffolds are isoreticular (which has been observed in MOF structures), and (3) four types of cages are embedded in the inter- scaffold space. Frameworks that result from this principle of structure expansion are called “embedded isoreticular” structures. The initial structure determination used structure factor amplitudes from the experimental RED data of ZSM-25. This dataset, however, has too low resolution to locate oxygen atoms, and it was necessary to add these oxygen atoms manually. The phase information of structure factors is much more important than the amplitude information, and inaccurate amplitudes can be tolerated when solving a structure as long as the phases are close to the correct values. Therefore, it is possible to adopt structure factor amplitudes from the PAU structure, in addition to phases. This strategy enabled us to re- determine the structure of ZSM-25 without using any experimental data. The amplitudes and phases of structure factors with a resolution lower than 1.0 Å were calculated from the idealized structural model of PAU (based on pure silica-form zeolite), and the 268 strongest reflections with E-values (normalized structure factors) greater than 1.2 were selected. The indices of all strong reflections from PAU were converted to the indices of the corresponding reflections from ZSM-25 using the following transformations: hZSM-25 = hPAU× aZSM-25/aPAU, kZSM-25 = kPAU× aZSM-25/aPAU, lZSM-25 = lPAU× aZSM- 25/aPAU. The 3D electron density map of ZSM-25 was calculated by taking the inverse Fourier transformation of the amplitudes and phases obtained from PAU. All of the 16 T atoms and 31 out of 40 oxygen atoms of ZSM-25 in the asymmetric unit were determined from the 3D map. The missing oxygen atoms were placed between the T-atoms by following rules of SiO4 tetrahedral geometry. The final model is a four-connected 3D framework, which has been assigned the three-letter code MWF by the IZA structure commission.

Figure 5.4 The (h k -h-k) planes calculated from the idealized pure-SiO2- form framework. Reflections in red and blue have phases of 0° and 180°, respectively. The red, green and blue circles correspond to d-spacings of 1.0 Å, 1.6 Å and 3.0 Å, respectively. Reprinted with permission from Ref. (99). Copyright © 2015, Rights Managed by Nature Publishing Group.

Figure 5.5 (a) Rietveld refinement plots of the as-made NaTEA-ZSM-25 (X-ray wavelength λ = 0.63248 Å). The inset intensities are scaled by a factor of six. The observed, calculated and difference curves are shown in blue, red and grey, respectively. The vertical bars indicate the positions of Bragg peaks. (b) The seven different cages: lta, d8r, pau, t-plg, t-oto, t-gsm and t- phi, found in ZSM-25. (c-d) The connectivity of the lta, d8r and pau cages in PAU (c) and ZSM-25 (d). (e-f) the 3D framework structure of PAU (e) and ZSM-25 (f) with t-plg, t-oto, t-gsm and t-phi cages embedded in the scaffolds. Reprinted with permission from Ref. (99). Copyright © 2015, Rights Managed by Nature Publishing Group.

5.2 Structure predictions of PST-20 (RHO-G5) and PST-25 (RHO-G6) The structures of this family can be predicted by expanding or contracting the two scaffolds and increasing or decreasing the filler material accordingly. For example, a hypothetical structure with a cubic unit cell of dimension a = 25 Å can be constructed by deleting a pair of d8r and pau cages from the scaffolds of the PAU structure. This hypothetical structure has been previously described (117, 118). A further structure, with the known RHO framework type with a cubic unit cell of dimension a = 15 Å, is obtained if a pair of d8r and pau cages is further deleted. This series of zeolite structures is therefore known as the “RHO family”. RHO, PAU and MWF are denoted as “RHO-G1” (the first generation in the RHO family), “RHO-G3” and “RHO-G4”, respectively. Members of the RHO family in low generations can be built manually, but members of higher RHO generations are too complicated for this to be possible. For example, the scaffolds of a hypothetical structure (with a = 55 Å) can be easily built by inserting a pair of d8r and pau cages into the scaffolds of MWF, but such a complicated structure model cannot be manually completed by adding the extra cages. The strong reflections approach can also be used to predict these complex structures. The structure factor amplitudes and phases with a resolution lower than 1.0 Å were calculated for the hypothetical structure (with a = 55 Å) from the idealized structure model of MWF. The 370 strongest reflections with E-values greater than 1.2 were selected. The indices of each strong reflection from RHO-G5 were calculated from the indices of the corresponding reflection from MWF (RHO-G4) according to: hRHO-G5= hRHO-G4× aRHO- G5/aRHO-G4, kRHO-G5= kRHO-G4× aRHO-G5/a RHO-G4, l RHO-G5= lRHO-G4× aRHO-G5/a RHO- G4. The 3D electron density map was calculated by taking the inverse Fourier transform of the amplitudes and phases obtained from the MWF structure. All of the 29 T atoms and 44 out of 69 oxygen atoms of RHO-G5 in the asymmetric unit were determined from the 3D map. The structure of the extremely complicated RHO-G6 was determined in the same way. RHO- G5 and RHO-G6 contain 7920 and 13104 atoms, respectively, in the unit cell (not including non-framework species), and it is a great challenge to determine these extremely complicated structures ab initio by conventional methods. Interestingly, other members of the RHO family than RHO-G1 and RHO-G2 are built up from the aforementioned seven cages and are small pore zeolites. Professor Suk Bong Hong’s research group has synthesized RHO-G5 and RHO-G6, by keeping the same OSDA and introducing new inorganic cations such as Sr2+, Ca2+ and Ba2+ into the gel. This approach was inspired by two known natural zeolites, Phillipsite (PHI) and Gismondine (GIS). The t- oto and t-phi cages are considered to be the building units of the Phillipsite structure, while t-gsm cages dominate the Gismondine structure. The num-

bers of t-oto, t-gsm and t-phi cages grow much faster than those of the other four types of cage. Furthermore, there are such alkaline-earth cations as extra cations in the structures of these two natural zeolites. The group named the two structures it had synthesized “PST-20” and “PST-25”. The RED data (Figure 5.1d-f) and the fit of the LeBail profile (Figure 5.6) confirm that PST-20 and PST-25 are present.

Figure 5.6 (a) Experimental synchrotron PXRD pattern of a mixture of PST- 20 and PST-25 and the simulated PXRD patterns of PST-20 and PST-25 are shown in blue, green and pink, respectively (λ = 1.4640 Å). (b) The plots of the LeBail profile using the two phases PST-20 and PST-25. The observed, calculated and difference curves are shown in blue, red and black, respectively. Green and pink vertical bars indicate the positions of Bragg peaks of PST-20 and PST-25, respectively. Reprinted with permission from Ref. (99). Copyright © 2015, Rights Managed by Nature Publishing Group.

5.3 Structure predictions of PST-26 (RHO-G7) and PST-28 (RHO-G8) (Paper V) The successful synthesis of PST-20 and PST-25 prompted us to explore much more complicated structures of the RHO family. Structural models of RHO-G7 (with a = 75 Å) and RHO-G8 (with a = 85 Å) were constructed. It may, in principle, have been possible to predict these two complicated structures by the strong reflections approach, but the electron density maps would

have been far too complex to be able to identify T and O atoms. For predicting higher generation zeolites in this family, we developed a new and more convenient approach, based on model building in real space. Since both scaffolds and embedded cages in the RHO zeolite family are repetitive, the higher generation members can be constructed from the lower ones. Taking RHO-G7 as an example, the model building procedures are: 1) A fragment of the RHO-G6 structure was taken out (Figure 5.7a). All of the atoms in the region 0 ≤ x ≤ 0.5; 0 ≤ y ≤ 0.5; 0 ≤ z ≤ 0.577 (= 0.5aRHO- G7/aRHO-G6); x ≤ z; y ≤ z; x ≤ y of the fragment were located. We note that aRHO-G6 (65 Å) and aRHO-G7 (75 Å) are the unit cell parameters of RHO-G6 and RHO-G7, respectively. 2) The fractional coordinates x, y, and z for all atoms in the fragment were scaled by a factor of aRHO-G6/aRHO-G7. 3) The scaled fractional coordinates x, y and z for RHO-G7 were converted into a crystallographic information file (CIF). A crystallographic program, “VESTA” (119) was used to remove duplicated atoms. 4) The procedure found a total of 72 T and 168 O atoms in the asymmetric unit (Figure 5.7b). A model of the RHO-G8 structure can be constructed by the same approach. Researchers at POSTECH have successfully synthesized these two much more complicated structures by careful control of the SiO2/Al2O3 and H2O/SiO2 ratios, and given the names “PST-26” and “PST-28” (120), respectively, to the compounds. The numbers of symmetry-independent T and O atoms in first 12 generations of the RHO family are listed in Table 5.2. Mathematical equations could be derived to describe the growth of the numbers of T and O atoms in each generation. It is of interest to note that odd and even generations of the RHO family follow two different equations for the number of T atoms, while the number of O atoms follows a single equation for all generations.

Number of T atoms in odd generations of the RHO family: 1 1 1 1 N  n3  n2  n  Eq.5.1 T odd 6 4 3 4 Number of T atoms in even generations of the RHO family: 1 1 1 N  n3  n2  n Eq.5.2 T even 6 4 3 Number of O atoms in all generations of the RHO family:

1 2 N  n3  n2  n Eq. 5.3 O 3 3

Figure 5.7 (a) Structural fragment of RHO-G6 used for the construction of the structural model of RHO-G7. (b) The structural model of RHO-G7 after removing the duplicated atoms.

Table 5.2 Numbers of the symmetry-independent T and O atoms in RHO- Gn members of the RHO family of zeolites (n = 1-12) (Reprinted with permission from Ref. (120). © 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim)

Generation Number of T atoms Number of O atoms RHO-G1 (zeolite Rho) 1 2 RHO-G2 3 8 RHO-G3 (Paulingite) 8 20 RHO-G4 (ZSM-25) 16 40 RHO-G5 (PST-20) 29 70 RHO-G6 (PST-25) 47 112 RHO-G7 (PST-26) 72 168 RHO-G8 (PST-28) 104 240 RHO-G9 145 330 RHO-G10 195 440 RHO-G11 256 572 RHO-G12 328 728

5.4 Conclusions In summary, this chapter presents a powerful method, the strong reflections approach. This method unifies the identification of a zeolite family, structure solution, and structure prediction. The method is unique in the way that the structure solution is based on a known structure, and experimental

data is not used. It can in this way overcome the problems described above that arise from the limited resolution and incomplete extent of the data collected. Furthermore, the method can direct the synthesis of new zeolites, to some extent. If two known zeolites are identified based on the similarity of the distribution of their strong reflections, it is probable that they belong to the same zeolite family. The other members of this family (probably higher generations) can then be predicted using the strong reflections approach. Suitable OSDAs for synthesizing the higher generations (predicted structures) of the family may be the same as those used for the lower generation (al- ready-known structures of this family). This is important information for researchers who are seeking to produce new zeolites.

6. Database mining of zeolite structures

The structures of ITQ-3 (FTC: ITE, Chapter 3) (104) and SSZ-52 (FTC: SFW, Chapter 2) (53) were elucidated by model building. The more complex structure of ZSM-25 (FTC: MWF) (99) was determined by the strong reflections approach, based on the known PAU structure (Chapter 5). These works emphasised how important the Database of Zeolite Structures is. The database contains much useful structural information of known zeolites, which can help us to (i) understand the relationships between zeolite synthe- ses and structures, and (ii) determine the structures of unknown zeolites. Chapter 6 summarizes some common and unique structural features of zeolites.

6.1 Characteristic structural information hinted by the unit cell dimensions Different materials with the same framework type may have different unit cells since the framework type in this database is independent of chemical composition. The length of the c-axis of AlPO-22 (FTC: AWW), for example, is twice that of the AWW framework type (pure Si form), since AlO4 and PO4 units alternate. Each framework was obtained from a DLS- refinement, assuming a SiO2 composition (116, 121). The distances of Si-O, O-O and Si-Si are typically around 1.61 Å, 2.63 Å and 3.07 Å, respectively, in all zeolites. This means that unit cell parameters may provide us important structural information of zeolite materials. This section explores the correla- tion between the unit cell parameters of the framework type and the zeolite structure.

6.1.1 5 Å One unit cell parameter is approximately 5 Å in 24 of the 231 framework types in the IZA Database of Zeolite Structures (highlighted in Table 6.1). We note also that the periodicities of double or single zig-zag chains are typically around 5 Å (two examples of which are the ATN (Figure 6.1a) and

IFO (Figure 6.1d) framework types), and that both structures have a 1D channel along this short axis (Figure 6.1b and 6.1e). Figure 6.1c shows that mirror symmetry perpendicular to this short axis implies that T atoms must be located on the mirror (Figure 6.1c and 6.1f). Three general rules can be drawn for the 24 framework types with a 5 Å unit cell dimension: 1) the framework has 1D channel along the 5 Å-axis; 2) if there is mirror symmetry perpendicular to the 5 Å-axis, all the T-atoms must be located on the mirror; 3) The framework is built of single or double zig-zag chains. The double and single zig-zag chains can be distinguished based on other information, such as IR and Raman spectra. It is obvious that the double zig-zag chain is composed of 4-rings, which can be identified from their IR (500-650 cm-1) and Raman (480-520 cm-1) spectra (115).

Figure 6.1 (a) Double zig-zag chain. (b) and (c) The ATN framework type viewed along [001] and [100] directions, respectively; (d) single zig-zag chain . (e) and (f) The IFO framework type viewed along the [010] and [001] directions, respectively. Both double and single zig-zag chains are highlighted in blue.

Table 6.1 The 24 framework types with a short axis of approximately 5 Å

FTC Space group a (Å) b (Å) c (Å) α (°) β (°) γ (°) m#

SSY Pmmn 5.26 22.58 13.98 90.00 90.00 90.00 yes

JBW Pmma 5.26 7.45 8.16 90.00 90.00 90.00 yes

MTT Pmmn 5.26 22.03 11.38 90.00 90.00 90.00 yes

MVY Pnnm 5.02 8.15 14.98 90.00 90.00 90.00 no

SFH Cmcm 5.25 34.32 21.52 90.00 90.00 90.00 yes

CAS Cmcm 5.26 14.13 17.23 90.00 90.00 90.00 yes

-CHI Pbcn 5.01 31.23 9.01 90.00 90.00 90.00 yes

NSI C2/m 14.13 5.25 8.93 90.00 105.37 90.00 yes

SFE P21/m 11.47 5.26 13.99 90.00 100.96 90.00 yes

CFI Imma 13.96 5.26 25.97 90.00 90.00 90.00 yes

MTW C2/m 25.55 5.26 12.12 90.00 109.31 90.00 yes

SFN C2/m 25.22 5.26 15.02 90.00 103.89 90.00 yes

ABW Imma 9.87 5.25 8.77 90.00 90.00 90.00 yes

IFO Pnnm 16.44 4.95 22.78 90.00 90.00 90.00 no

ATN I4/mmm 13.07 13.07 5.26 90.00 90.00 90.00 yes

ATO R-3m 20.91 20.91 5.06 90.00 90.00 120.00 no

ATS Cmcm 13.16 21.58 5.26 90.00 90.00 90.00 yes

BCT I4/mmm 8.95 8.95 5.26 90.00 90.00 90.00 yes

CAN P63/mmc 12.49 12.49 5.25 90.00 90.00 120.00 yes

BIK Cmcm 7.54 16.22 5.26 90.00 90.00 90.00 yes

GON Cmmm 16.90 20.40 5.26 90.00 90.00 90.00 yes

OSI I4/mmm 18.51 18.51 5.27 90.00 90.00 90.00 yes

TON Cmcm 14.10 17.84 5.26 90.00 90.00 90.00 yes

VET P-4 13.05 13.05 4.95 90.00 90.00 90.00 no

# There is the mirror symmetry perpendicular to this short axis.

6.1.2 7.5 Å Table 6.2 lists the 24 framework types with at least one unit cell parameter of 7.5 Å. Structural features of dimension approximately 7.5 Å are much more interesting and more complex than those of dimension approximately 5 Å. The periodicity of double saw chains, which are composed of three edge- sharing single 4-rings (denoted “4, 4-double saw chains”, Type A in Figure

6.2a), is typically approximately 7.5 Å. Two further derivatives of double saw chains, including the 8, 8-double saw chains (Type B in Figure 6.2e) and 4, 6-double saw chains (Type C in Figure 6.2h), are observed among the known zeolite structures. The unit cell dimensions of these derivatives can deviate by up to 0.3 Å from the typical distance of 7.5 Å, such as in the SFF framework type. It might be possible to obtain the channel systems of a zeolite framework from the unit cell parameters and symmetry. The presence of an 8-ring channel normal to an axis requires mirror symmetry perpendicular to this axis. Consider the LTF, MFS and LAU framework types as examples. The space group of LTF is P63/mmc and the length of the c-axis is approximately 7.60 Å. A double saw chain is present in the LTF framework along the [001] direction (Figure 6.2b). Mirror symmetry perpendicular to the c-axis generates 8-ring pore openings in the bc plane (Figure 6.2c). Other framework types summarized in Table 6.2, such as EON, MOZ, OFF, -WEN, MAZ and LTL, also have the same double saw chain and 8-ring openings perpendicular to the 7.5 Å-axis. The space group of the MFS framework type is Imm2, and the length of the a-axis is 7.54 Å. 8, 8-double saw chains are present in the MFS framework along its a-axis (Figure 6.2f). Mirror symmetry perpendicular to the a- axis generates 8-ring pore openings in the ac plane (Figure 6.2g). Other framework types, including CDO, RWR, DAC, FER and MOR, also follow this rule. The 4, 6-double saw chain present in the LAU framework type (Figure 2i) lacks such mirror symmetry, and thus a 1D channel only is present along the c axis (no pore opening is present in the bc or ac plane, Figure 6.2j). Frame- work types that contain a 4, 6-double saw chain, such as IFR, SZR, MTF, SFF, RTE and STF, follow this rule. In summary, mirror symmetry perpendicular to an axis of approximate length 7.5 Å suggests that 8-ring pore openings are present normal to this axis. Some framework types, however, such as RRO, BRE and HEU, do not follow this rule, and 8-ring pore openings may be present even if there is no mirror symmetry perpendicular to the 7.5 Å-axis. These three framework types consist of the same bre building unit. Thus, the reason that these three types break the rule might be that they do not contain suitable double chains mentioned above.

Table 6.2 The 24 framework types with one axis of approximately 7.5 Å. Space 8- Types of FTC a (Å) b (Å) c (Å) α (°) β (°) γ (°) m group rings chain

BIK Cmcm 7.54 16.22 5.26 90.00 90.00 90.00 yes yes #

RRO P2/c 7.41 8.64 17.18 90.00 113.68 90.00 yes no *

MFS Imm2 7.54 14.39 19.02 90.00 90.00 90.00 yes yes B

CDO Cmcm 7.56 18.72 14.10 90.00 90.00 90.00 yes yes B

EON Pmmn 7.57 18.15 25.93 90.00 90.00 90.00 yes yes A

RWR I41/amd 7.81 7.81 27.35 90.00 90.00 90.00 yes yes B

DAC C2/m 18.57 7.54 10.38 90.00 108.92 90.00 yes yes B

IFR C2/m 18.63 13.44 7.63 90.00 102.32 90.00 no no C

SZR Cmmm 18.87 14.40 7.51 90.00 90.00 90.00 yes yes C

FER Immm 19.02 14.30 7.54 90.00 90.00 90.00 yes yes B

LTF P63/mmc 31.17 31.17 7.60 90.00 90.00 120.00 yes yes A

MOZ P6/mmm 31.20 31.20 7.55 90.00 90.00 120.00 yes yes A

BRE P21/m 6.76 17.09 7.60 90.00 95.83 90.00 yes no *

MTF C2/m 9.63 30.39 7.25 90.00 90.45 90.00 no no C

SFF P21/m 11.45 21.70 7.23 90.00 93.15 90.00 no no C

OFF P-6m2 13.06 13.06 7.57 90.00 90.00 120.00 yes yes A

-WEN P-62m 13.59 13.59 7.56 90.00 90.00 120.00 yes yes A

RTE C2/m 14.10 13.67 7.43 90.00 102.42 90.00 no no C

STF C2/m 14.10 18.21 7.48 90.00 98.99 90.00 no no C

LAU C2/m 14.59 12.88 7.61 90.00 111.16 90.00 no no C

HEU C2/m 17.52 17.64 7.40 90.00 116.10 90.00 yes no *

MAZ P63/mmc 18.10 18.10 7.62 90.00 90.00 120.00 yes yes A

LTL P6/mmm 18.13 18.13 7.57 90.00 90.00 120.00 yes yes A

MOR Cmcm 18.26 20.53 7.54 90.00 90.00 90.00 yes yes B # There is a short axis being 5 Å in the BIK framework type. It belongs to Section 6.1.1. * The framework types don’t follow the rule discussed in Section 6.1.2, but they contain the same bre building unit.

Figure 6.2 (a) 4, 4-double saw chain. (b-c) The LTF framework type viewed along [001] and [100] directions, respectively. (e) 8, 8-double saw chain; (f- g) MFS framework type viewed along the [001] and [010] directions, respectively. (e) 4, 6-double saw chain; (f-g) The LAU framework type viewed the along [001] and [100] directions, respectively.

6.1.3 10 Å Table 6.3 lists the nine framework types with double crankshaft chains (Fig- ure 6.3a). The length of the repeating unit is approximately 10 Å. Criteria similar to those described previously can be used to determine whether 8-

ring pore openings are present normal to the unit cell edge of length 10 Å. Consider the PHI framework type as an example: its space group is Cmcm, and the length of the a-axis is 9.89 Å. Double crankshaft chains are present along the a-axis (Figure 6.3b). The symmetry operations of its space group include mirror symmetry normal to the a-axis, which generates 8-ring pore openings in the ac or ab plane (Figure 6.3c). All the nine framework types listed in Table 6.3 follow this rule.

Figure 6.3 (a) Double crankshaft chain in the PHI framework type; (b) and (c) The PHI framework type viewed along the [100] and [001] directions, respectively.

Table 6.3 The nine framework types with one axis of approximate 10 Å.

Space 8- FTC a (Å) b (Å) c (Å) α (°) β (°) γ (°) m group rings APC Cmce 8.99 19.36 10.39 90.00 90.00 90.00 yes yes

ATT Pmma 9.98 7.51 9.37 90.00 90.00 90.00 yes yes

AWO Cmce 9.10 15.04 19.24 90.00 90.00 90.00 yes yes

PHI Cmcm 9.89 14.06 14.05 90.00 90.00 90.00 yes yes

SIV Cmcm 9.88 14.08 28.13 90.00 90.00 90.00 yes yes

GIS I41/amd 9.80 9.80 10.16 90.00 90.00 90.00 yes yes

UEI Fmm2 19.46 9.35 15.11 90.00 90.00 90.00 yes yes

GME P63/mmc 13.67 13.67 9.85 90.00 90.00 120.00 yes yes

MER I4/mmm 14.01 14.01 9.95 90.00 90.00 90.00 yes yes

6.1.4 12.7Å Table 6.4 lists eight zeolite structures with one unit cell dimension of approximate length 12.7 Å, and these have been summarized also by Allen W. Burton et al. (122). These authors mentioned that the repeating unit may be two consecutive n26n cages followed by an n24n cage (where n = 6 for SSF, n = 5 for SFG, and n = 4 for UWY, ISV, ITH, BEC, IWW and IWR), as shown in Figure 6.4.

Table 6.4 The eight framework types with one axis around 12.7 Å.

FTC Space group a (Å) b (Å) c (Å) α (°) β (°) γ (°)

SSF P6/mmm 17.21 17.21 12.80 90.00 90.00 120.00

SFG Pmma 25.53 12.57 13.07 90.00 90.00 90.00

ISV P42/mmc 12.87 12.87 25.67 90.00 90.00 90.00

ITH Amm2 12.57 11.66 21.93 90.00 90.00 90.00

BEC P42/mmc 12.77 12.77 12.98 90.00 90.00 90.00

IWW Pbam 41.69 12.71 12.71 90.00 90.00 90.00

IWR Cmmm 21.23 13.30 12.68 90.00 90.00 90.00

UWY Pmmm 25.11 12.73 11.51 90.00 90.00 90.00

Figure 6.4 The repeated unit (12.7 Å) in the framework types of SSF, SFG, UWY, ISV, ITH, BEC, IWW and IWR.

6.1.5 14 Å Zeolite structures in which one of the unit cell dimensions is approximately 14 Å may contain 5-rings. These 5-rings may be further assembled to create either a repeating chain of length 14 Å (Table 6.5) or a condensed structure without accessible channels or cavities (such as the MEP framework type). A low OH/Si and a high Si/Al ratio during synthesis promote the formation of 5-rings. The chain structure can be distinguished by the characteristic adsorption near 1200 cm-1 in the mid-infrared spectrum (1500-400 cm-1) (123).

Table 6.5 The 20 framework types with one axis of approximate 14 Å. FTC Space a (Å) b (Å) c (Å) α (°) β (°) γ (°) group CFI Imma 13.96 5.26 25.97 90.00 90.00 90.00

EEI Fmmm 13.88 35.75 22.49 90.00 90.00 90.00

EUO Cmme 13.90 22.86 20.58 90.00 90.00 90.00

IHW Cmce 13.75 24.07 18.33 90.00 90.00 90.00

IMF Cmcm 14.30 56.79 20.29 90.00 90.00 90.00

MEP Pm-3n 13.71 13.71 13.71 90.00 90.00 90.00

NSI C2/m 14.13 5.25 8.93 90.00 90.00 90.00

SFS P21/m 14.01 20.03 12.49 90.00 106.08 90.00

TON Cmcm 14.11 17.84 5.26 90.00 90.00 90.00

CAS Cmcm 5.26 14.13 17.23 90.00 90.00 90.00

NES Fmmm 26.06 13.89 22.86 90.00 90.00 90.00

FER Immm 19.02 14.30 7.54 90.00 90.00 90.00

OKO C2/m 24.06 13.83 12.35 90.00 109.13 90.00

PCR C2/m 20.14 14.07 12.52 90.00 115.65 90.00

-SVR Cc 20.41 13.30 20.16 90.00 102.33 90.00

SZR Cmmm 18.87 14.40 7.51 90.00 90.00 90.00

MFI Pnma 20.09 19.74 13.14 90.00 90.00 90.00

NON Fmmm 22.86 15.66 13.94 90.00 90.00 90.00

CDO Cmcm 7.56 18.72 14.10 90.00 90.00 90.00

MEL I-4m2 20.27 20.27 13.46 90.00 90.00 90.00

6.1.6 20 Å A large number of zeolites can be considered to be an assembly of certain layers based on different symmetry operations, such as reflection in a mirror plane or inversion through a center. The overall thickness of the building layers (including oxygen atoms that form bridges between the layers) is approximately 20 Å. The well-known MFI and MEL framework types, for example, share common building layers. The type materials for the MFI and MEL framework types are ZSM-5 and ZSM-11, respectively. Both of these are high-silica zeolites and have 5-rings in their structures. Thus, we expect one of the unit cell parameters to be around 14 Å, corresponding to four 5- ring chains. The MFI or MEL framework is formed by layers that are related through an inversion center or mirror plane, respectively. In both cases, the thickness of two building layers is approximately 20 Å (Figure 6.5). The thickness of the layers deviates slightly from 20 Å in some cases, which may provide additional information about the channel system. This point will be addressed separately in Section 6.3 (124).

Figure 6.5 Projections of MFI viewed along the [010] direction and MEL along the [010] direction.

6.2 The ABC-6 family The periodic building unit in the ABC-6 family is a hexagonal array of non- connected planar 6-rings, which are related by translations along the a- or b- axis (Figure 6.6a). These adjacent 6-rings are connected by tilted single 4- rings, to form a fully 4-connected framework (Figure 6.6b). Table 6.6 lists the 22 framework types that belong to the ABC-6 family. The distinctive structural features of this family can be summarized as follows:

Figure 6.6 (a) the translation of 6-rings along the a- or b-axis; (b) adjacent 6-rings connected by tilted 4-rings. (1) The structures in the ABC-6 family can be categorized into three sub- groups, based on the length of the a- (or b-) axis of the hexagonal unit cell. Each subgroup has its unique structural features (53). The CAN, LOS, LIO, AFG, FRA, TOL, MAR, FAR and GIU framework types, for example, have an a (or b) dimension in the range from 12.25 Å to 12.70 Å, and have building units that are only single 6-rings. Other framework types, (GME, CHA, AFX, AFT and SFW) have an a (or b) dimension between 13.65 and 13.75 Å, and consist solely of double 6-rings. Finally, other members of the ABC-6 family have an a (or b) dimension between 12.85 Å and 13.20 Å, and these members have mixed building units of both single and double 6-rings. (2) The number of repeated layers (N) depends on the length of the c-axis of the hexagonal unit cell: N = c/2.5. (3) Some members have no C stacking sequence, and this gives rise to a 12- ring channel along the c-axis. Examples are the CAN, OFF and GME framework types. (4) The zeolite framework of members with no C stacking sequence in which the building unit is composed of solely single 6-rings have no accessible channels. The maximum ring size is 6 in such cases. (5) Members in which a C stacking sequence is present with double 6-rings within the framework are small-pore zeolites. The maximum ring size in

such cases is 8. Applications of small pore zeolites are described in Chapter 1. (6) The phases of the structure factors of some reflections in the diffraction patterns of members of the ABC-6 family are fixed by symmetry constraints. Figure 6.7 shows structure factor amplitudes and phases of hk0 reflections calculated from these frameworks. The phases of structure factors with indices 110, 220, 330, and 440 are constrained to be 180°, 180°, 0° and 0°, respectively. Section 3.1.3 showed that it is important to determine the phases of structure factors in order to determine a structure. The fact that these phases of the structure factors are constrained may be useful when determining the structures of unknown members of the ABC-6 family. Further, it may be possible to gain information about intrinsic phase relationships in other zeolite families.

Table 6.6 ABC-6 family.

Type Space Stacking sequence Code Ring Sizes Building material Group (Repeating layer) units Cancrinite CAN P6 /mmc 12 6 4 AB(A)....(2) 3 Single 6- Sodalite SOD Im-3m 6 4 ABC(A)...(3) rings only Losod LOS P63/mmc 6 4 ABAC(A)...(4) Liottite LIO P-6m2 6 4 ABACAC(A)...(6) 12.25Å- 12.70Å Afganite AFG P63/mmc 6 4 ABABACAC(A)...(8) Franzinite FRA P-3m1 6 4 ABCABACABC(A)...(11) Tounkite TOL P-3m1 6 4 CACACBCBCACB(C)...(12)

Marinellite MAR P63/mmc 6 4 ABCBCBACBCBC(A)...(12)

Farneseite FAR P63/mmc 6 4 ABCABABACBACAC(A)...(14)

Giuseppettite GIU P63/mmc 6 4 ABABABACBABABABC(A)...(16) Offretite OFF P-6m2 12 8 6 4 AAB(A)...(3) Double 6- ZnAlPO-57 AFV P-3m1 8 6 4 AABCC(A)...(5) rings and Erionite ERI P63/mmc 8 6 4 AABAAC(A)...(6) single 6-

TMA-E(AB) EAB P63/mmc 8 6 4 AABCCB(A)...(6) rings ZnAlPO-59 AVL P-3m1 8 6 4 ABBACCA(A)....(7) 12.85Å- Levyne LEV R-3m 8 6 4 AABCCABBC(A)...(9) 13.20Å STA-2 SAT R-3m 8 6 4 AABABBCBCCAC(A)...(12)

Gmelinite GME P63/mmc 12 8 6 4 AABB(A)...(4) Double 6- Chabazite R-3m 8 6 4 AABBCC(A)...(6) CHA rings only SAPO-56 AFX P6 /mmc 8 6 4 AABBCCBB(A)...(8) 3 13.65Å- AlPO-52 AFT P63/mmc 8 6 4 AABBCCBBAACC(A)...(12) 13.75Å SSZ-52 SFW R-3m 8 6 4 AABBAABBCCBBCCAACC(A)...(18)

Fig ure 6.7 Comparison of structure factor amplitudes and phases of hk0 reflections calculated from the structures of different zeolite framework types . The red, green and blue circles correspond to d-spacings of 1.0 Å, 1.6 Å and 3.0 Å, respectively. The frameworks are idealized in the pure SiO2 forms. Reflections in red and blue have phases of 0° and 180°, respectively. The indices are given only for the LOS framework, for clarity.

6 .3 The butterfly family (Paper VI)

Zeolite materials with framework types FAU, FER, MOR, MFI and LTA are widely used in industrial applications. The MFI and FER framework types have similar nets, containing the well-known ‘‘butterfly’’ configuration with a 6-ring (the body of the butterfly) surrounded by four 5-rings (the wings) (Figure 6.8) (124). The chain of edge-sharing 5-rings can be considered to be the main building unit, and generates alternating 6-rings and 10-

rings in the intermediate chains. This is a unique structural feature of high- silica zeolites, and such chains can be identified using IR spectroscopy. The 2 2 2 vertex symbol for this layer is [5.6.10]2 [5 .10] [5 .6] [5 .10]2. This family of zeolite frameworks is known as the “butterfly family”. Ten framework types built solely from the butterfly net have been identified. These ten framework types are *MRE (type material: ZSM-48), FER (ferrierite), MEL (ZSM-11), SZR (SUZ-4), MFS (ZSM-57), MFI (ZSM-5), TUN (TNU-9), IMF (IM-5), BOG (boggsite) and TON (theta-1). Table 6.7 lists these framework types, with their unit cell settings reconfigured to make them easy to compare. The butterfly net of all ten frameworks is located in the ab plane with a ≈ 14 Å and b ≈ 20 Å. Table 6.7 lists the space groups, channel ring sizes and channel systems of these frameworks. The same unit cell configuration has been applied to the TUN and IMF framework types, although the a parameter is doubled in TUN and the b parameter is tripled in IMF (highlighted in Table 6.7).

Table 6.7 Re-configured unit cell parameters (a, b and c), space groups and channel ring sizes for the ten members of the butterfly family of zeolites (Reprinted with permission from Ref. (124). Copyright © 2015, Walter de Gruyter GmbH).

Channel ring Δa Δb Space Channel Framework a (Å) b (Å) c (Å) size along a, (Å)1 (Å)1 group dimension b, c *MRE 14.562 20.314 8.2570 ̶ ̶ Imcm ̶ , ̶ , 10 1D

BOG 12.669 20.014 23.580 1.893 0.300 Ibmm ̶ , 12, 10 3D

MFI 13.142 20.090 19.738 1.420 0.224 Pbnm 10, 10,10 3D

MEL 13.459 20.270 20.270 1.103 0.044 I-4m22 10, 10,10 3D

TUN 27.845 19.597 20.015 0.640 0.7543 B112/m4 ̶ , ̶ , 105 3D

TON 14.105 17.842 5.256 0.457 2.472 Cmcm ̶ , ̶ , 10 1D

IMF 14.296 56.788 20.290 0.266 1.3856 Cmcm 10, ̶ , 10 3D

FER 14.303 19.018 7.541 0.259 1.296 Immm 8, ̶ , 10 2D

MFS 14.388 19.016 7.542 0.174 1.298 Im2m 8, ̶ , 10 2D

SZR 14.401 18.870 7.5140 0.161 1.444 Cmmm 8, ̶ , 10 3D

1 Δa and Δb are the differences between a and b parameters and those of *MRE. The space groups have been changed to reflect the new unit cell settings, with the exception of the tetragonal unit cell used for MEL. 2 The original space group is given here. The -4 symmetry operation is along the a-axis in

the new unit cell. 3 This value has been calculated using a/2. 4 The unique axis is the c-axis and  = 93.2°. The full H-M symbol is given here for clarity. 5 The 10-ring channels are along the [1-10] direction. 6 This value has been calculated using b/3.

The butterfly nets in these ten framework types are similar, with small geometric differences between them (Figure 6.9). They are built by stacking the butterfly nets along the c-axis, based either on mirror symmetry or on inversion centre symmetry. The T atoms in the net are three-connected, with the forth connection pointing either up (U) or down (D) (Figure 6.9). The up and down configuration differs from structure to structure, and thus the connections of neighbouring layers differ also from structure to structure. Fur- ther, also the 3D channel systems of the frameworks in the butterfly family can differ. The *MRE framework has the longest a- and b-axes (a = 14.562 Å and b = 20.314 Å), and we have used this framework as a reference when compar-

ing members of the family. Table 6.7 presents the differences between the the a and b parameters of the other frameworks in the family and those of *MRE. The a parameter of BOG, for example, is 1.893 Å shorter than that of *MRE, which can be seen clearly in Figure 6.10A, while the b parameter of BOG is close to that of *MRE. The ab plane of *MRE is so flat that the planes are close-stacked along the c-axis, and no pore opening is formed in the ac or bc plane. The ab plane of BOG is much more rippled than that of *MRE. Neighbouring layers are related by mirror symmetry perpendicular to the c-axis, which generates 12-ring pore openings along the b-axis (Figure 6.11). The comparison between *MRE and BOG allows the relationship between the unit cell dimensions and the channel systems to be summarized for all frameworks in the butterfly family. If the a parameter of a framework is shorter than that of the reference member (*MRE), while the b parameter is approximately the same, channels along the perpendicular axis (the b-axis) will be formed. The BOG, MFI and MEL framework types exemplify this rule (Figure 6.10A and Figure 6.11). In contrast, the b parameters of SZR, MFS, FER and IMF differ consid- erably from that of the reference member (up to 1.444 Å in difference) (Fig- ure 6.10B). We expect, therefore, that channels running parallel to the b-axis will be formed. Indeed, SZR, MFS and FER have 8-ring pore openings along the b-axis, while IMF has 10-ring channels (Figure 6.11). Both the a and b parameters of the TUN framework type are shorter than those of the reference member (*MRE), by 0.640 Å and 0.717 Å, respectively. We expect that it will have channels running diagonally in the ab plane, as is, indeed, the case (Figure 6.11). The b parameter of the TON framework type differs most from that of the reference member (*MRE), and we expect a channel system along the a- axis. However, a 1D channel forms only along the c-axis, and no pore openings along the a- or b-axes. The c parameter of this framework is particularly short (5.256 Å), which is the reason of the 1D channel (Section 6.1.1).

Figure 6.9 The butterfly layers extracted from the ten members of the butterfly family viewed perpendicular to the layer. The undulation of the layers leads to the channels (the grey stripes). Only the T-T connections (T = Si, Al) are shown for clarity. The T atoms pointing up and down are shown in blue and gold, respectively. (Reprinted with permission from Ref. (124). Copy- right © 2015, Walter de Gruyter GmbH).

Figure 6.10 the butterfly layers viewed along the b-axis (A) and the a-axis (B). The grey spheres indicate projections of the channels, which result from the undulation of the layers. Only the T-T connections (T = Si, Al) are shown for clarity. The T atoms pointing up and down are shown in blue and gold, respectively. The layers are shown on the same scale. (Reprinted with permission from Ref. (124). Copyright © 2015, Walter de Gruyter GmbH).

Figure 6.11 *MRE, TON, IMF, FER, MFS and SZR frameworks viewed along the a-axis. BOG, MFI and MEL viewed along the b-axis. TUN framework viewed along [1-10]. The c-axis is perpendicular to all the structures. Only the T-T connections (T = Si, Al) are shown, for clarity. One butterfly layer is highlighted in each structure. The T atoms pointing up and down are shown in blue and gold, respectively. (Reprinted with permission from Ref. (124). Copyright © 2015, Walter de Gruyter GmbH).

6.4 Conclusions This chapter has described the important and interesting relationships between unit cell dimensions and the structural features of known zeolite frameworks. The information presented here may be embedded into ad- vanced software with the ability to solve zeolite structures.

7. Sammanfattning

Zeoliter är porösa kristallina aluminiumsilikater vars hålrum är lika stora som små molekyler. Befintliga zeoliter används idag i stor utsträckning i industriella processer, till exampel gasseparation, jonbyte och katalytisk omvandling av organisk molekyler. Huvudanledningen till att zeoliterna har blivit så framgångsrika är deras väldefinierade porsystem som kan fungera som en sil på molekylnivå. För att kunna förstå funktionen och utveckla nyå zeoliter måste man veta hur dessa struktuer är uppbyggda. Konventionella metoder för struktur- bestämning, så som röntgendiffraktion, har tillämpats för att bestämma strukturen hos många olika typer av material. Metoden kräver dock kristaller som är relativt stora. Röntgendiffraktion från pulverprover är en annan metod som också kan användas för strukturbestämning. Den är dock komplicerad och begränsas bl. a. av överlappning av diffraktionstoppar, låg kristallinitet och föroreningars bidrag till diffraktionsmönstret. Elektron- kristallografi, dvs elektrondiffraktion och högupplösande transmissions- elektronmikroskopi, kan lösa dessa problem. I vår grupp har vi nyligen utvecklat en metod för automatiserad insamling och analys av tre- dimensionella elektrondiffraktionsdata, som kallas rotations elektrondiffraktion med förkortningen RED. Målet för denna avhandling är att bestämma strukturer av nya zeoliter och zeolit-relaterade material genom att använda RED metoden. Ett annat mål är att utveckla metoder att förutsäga nya strukturer med skräddarsydda egenskaper. Vi har utvecklat en metod för att förutsäga hypotetiska strukturer som baserar sig på redan kända zeoliter. Denna metod ger en länk mellan förutsägande av strukturer och syntes av nya material. Dessutom kan vi studera hur molekyler och joner interagerar med själva zeolitstrukturen genom Rietveldförfining. Genom att studera uppbyggnaden av strukturen underlättas designen av anpassade syntesvägar. De huvudsakliga resultaten i denna avhandling kan sammanfattas i följande punkter: 1) Att lösa strukturer av två silikater COE-3 och COE-4, en fluoraluminatfosfat EMM-9 och en borsilikat zeolit EMM-26 från RED data. 2) Att använda Rietveldförfining för att bekräfta strukturmodeller vi fick med hjälp av RED data och hitta de molekyler och joner som finns inuti hålrummen.

3) Att utveckla en ny metod (”strong reflections approach”) för att identifiera en ny familj av zeolitstrukturer. Denna metod har tillämpats för att bestämma strukturen av en 34 år gamml aluminosilikat ZSM-25 och att förutsäga flera nya och besläktade strukturer (PST-20, PST-25, PST-26 och PST-28). Metoden förväntas även kunna vägleda framtida synteser av nya material. 4) Att sammanfatta vanliga strukturelement baserat på kända strukturer. Detta är ett hjälpmedel som kan användas vid strukturbestämning av nya zeolitmaterial, samt ge inspiration för att syntetisera nya material med liknande strukturelement.

8. Future perspective

The development of new 3D ED methods including the automated diffraction tomography (ADT) and rotation electron diffraction (RED) has pushed the limit of structure determination by X-ray diffraction to nano- and submicrometer sized crystals. A large number of structures that could not be solved by X-ray diffraction have been determined using the 3D ED data. However, due to dynamical effects, the current structure refinement based on kinematic electron diffraction is not good enough to get accurate atomic positions, and locate OSDAs or guest water molecules in the pores of zeolites. Today, we still rely on PXRD data and Rietveld refinement to complete the structural model obtained from RED data and to confirm the model. However, Rietveld refinement is only possible when the quality of PXRD data is high and the structures are not too complex. For samples with poor crystallinity and containing disorders and impurities, it is very challenging to obtain an accurate model from PXRD data by Rietveld refinement. So it is necessary to further develop the 3D ED technique and make it as a stand- alone structure determination technique as single crystal X-ray diffraction. One possible solution is to take into account dynamical effects in the structure refinement, as has been shown by Lukas Palatinus and co-workers (125). Although many zeolite structures have been predicted, rational synthesis of predicted zeolites is still challenging. We have shown for the first time that a series of related zeolite structures can be predicted using the strong reflections approach and they could be finally synthesized. The strong reflections approach may link the structure prediction with the target synthesis. Currently there are 231 framework types in the Database of Zeolite Struc- tures. These framework types might be grouped based on the similarity of the intensity distribution of reflections in reciprocal space. The common structural coding of zeolites in the same group (family) may be used for identification and prediction of new structures within this zeolite family. According to the synthesis recipes of known zeolites of this family, it is highly possible to synthesize predicted zeolite structures. Hopefully, this unique method will make a big step towards the rational design of zeolites. Phasing of reflections is a crucial step in structure determination. As described in Chapter 6, we observed that some reflections in the ABC-6 family have fixed structure factor phases. Are there any other phase relations for certain reflections exiting in zeolite structures? We hope to find more rela-

tions among the zeolite structures and use them in the future to structure determination and prediction of novel zeolites. This thesis focuses on the structural investigation of unknown zeolites using various crystallographic and diffraction methods. However, for industrial applications, more research has been concentrated on the modification of known zeolites. In the future, I hope to integrate my crystallographic back- ground with various spectroscopic techniques such as NMR and IR in order to have deep understandings of these known and widely-used zeolite catalysts down to the atomic level.

9. Acknowledgements

First of all, I would like to thank my supervisor Prof. Xiaodong Zou for your patient training and everlasting support during my Ph.D. study. With your help, my Ph.D. study became very smooth. You are very strict with our scientific work, but quite nice and open to our personal life. During the past four years, I have gained not only scientific knowledge about electron crystallography from you, but also how to set up my own network in academia and communicate with our collaborators. Your group is a very nice place for ambitious young students to study and grow up in academia. I also want to thank my co-supervisor, Assistant Prof. Wei Wan, for your solid knowledge related to TEM. You are always helpful when I am ex- hausted by Xiaodong’s questions or comments. You are a man of few words, but one never knows how much knowledge has been accumulated in your brain. My great appreciations are given to our excellent collaborators: Dr. Allen W. Burton, Dr. Karl G. Strohmaier, Dr. Guang Cao, Dr. Mobae Afeworki and co-workers at ExxonMobil in Annandale, Dr. Jiho Shin, Prof. Suk Bong Hong and co-workers at POSTECH, Dr. Alex G. Greenaway and Prof. Paul A. Wright at University of St Andrews, Dr. Paul A. Cox at University of Portsmouth, Prof. Hermann Gies at the Ruhr-Universität, Prof. Feng-Shou Xiao at Zhejiang University, Prof. Lynne B. McCusker and Dr. Christian Baerlocher at ETH-Zurich, Dr. Leifeng Liu, Dr. Jie Su, Dr. Hongyi Xu, Dr. Chao Xu and Dr. Duan Li at MMK. Thank Prof. Lynne B. McCusker, Dr. Christian Baerlocher, Dr. Yifeng Yun, Dr. Tom Willhammar, Dr. Hong Chen, Dr. Dan Xie, Dr. Stef Smeets and Prof. Junliang Sun for your training about crystallography. Thank Dr. Hongyi Xu, Mr. Yanhang Ma, and Dr. Yi Zhang for your help in the TEM operation. Thank Dr. Haoquan Zheng for your daily conversations about recent publications. Thank Mr. Junzhong Lin, Mr. Yunxiang Li, Dr. Jiho Shin and Dr. Changjiu Xia for sharing your knowledge about the synthesis of zeolites. Thank Mr. Yunchen Wang for your nice program about the calculation of normalized structure factors. I thank all the other members in Xiaodong’s group for your help during my Ph.D. study. Thank Prof. Mats Johnsson for his supervision of my study plan and strong recommendation letter.

Thank Prof. Osamu Terasaki for your so called simple questions which always inspire me. And also thank you for your strong recommendation letter during my job application. Thank the “language police” Prof. Sven Hovmöller for improving my English. I also would like to thank Prof. Xianhui Bu in California State University for your nice explanations about the chirality in MOFs and also the warm help when I was in trouble during my Master study. Thank you for your nice recommendation letters during my Ph.D. application and job application. I would like to thank Prof. Zhongmin Liu and Prof. Peng Tian for your help during my internship in Dalian. I am grateful to Dr. Cheuk-Wai Tai and Dr. Kjell Jansson for always keeping our microscopes in the best conditions. Thank Dr. Lars Eriksson and Dr. Jekabs Grins for taking care of the powder diffractometers, which made my projects smoothly. Thank Prof. Sven Hovmöller, Prof. Xiaodong Zou, Dr. George Farrants, Dr. Hongyi Xu, Dr. Stef Smeets and Prof. Lynne B. McCusker for reading my thesis and correcting the language. Thank Tom Willhammar for translat- ing the Swedish summary. I am so grateful for all of the administrative staff, Gunnar, Camilla, Dan- iel, Tatiana, Helmi, Pia, Ann, Hanna, Ann-Britt, and Anita and other MMK members, you made my study here very convenient. Thank the support from the Berzelii Center EXSELENT funded by Swe- dish Research Council (VR) and the Swedish Governmental Agency for Innovation Systems (VINNOVA). Thank the Knut and Alice Wallenberg Foundation through a grant for purchasing the TEMs and the project grant 3DEM-NATUR. Thank the ESRF, Grenoble for the synchrotron X-ray beam time. Thank the Chinses government for the award for outstanding-financed Students abroad. To football players in MMK, German, Ken, Zoltan, Arnaud, Haoquan, Junzhong, Dickson, Yanhang, Leifeng, Ge, Yunxiang, Qingxia and other players on the pitch, we had a good time playing football together. Also to other friends in Sweden, Dariusz, Alexandra, Hani, Hao, Diana, Ahmed, Elina, Magda, Taimin, Guang, Yuan, Qingpeng, Fei, Yang, Ji, Ning, Yajuan, Yonglei, Shichao, Bin, Jianfeng, and Yongsheng, you made my life in Sweden so happy and interesting. Finally, I would like to thank my parents and my beloved wife Jing Wang for their sincere help and understanding in my life. I am sure we will have a happy life in Dalian.

10. References

1. J. Rouquerol et al., Recommendations for the characterization of porous solids (Technical Report). Pure Appl. Chem. 66, 1739–1758 (2009).

2. A. G. Slater, A. I. Cooper, Function-led design of new porous materials. Science. 348, 8075 (2015).

3. H. Deng et al., Large-Pore Apertures in a Series of Metal-Organic Frameworks. Science. 336, 1018–1023 (2012).

4. G. H. V. Bertrand, V. K. Michaelis, T.-C. Ong, R. G. Griffin, M. Dincă, Thiophene-based covalent organic frameworks. Proc. Natl. Acad. Sci. 110, 4923–4928 (2013).

5. A. F. Cronstedt, Ron och beskriting om en obekant bärgant som kallas zeolites. Akad. Handl. 18, 120–130 (1756).

6. H. de St Claire Deville, Comptes Rendus Acad. Sci. 54, 324 (1862).

7. G. T. Kerr, Chemistry of Crystalline Aluminosilicates. III. The Synthe- sis and Properties of Zeolite ZK-5. Inorg. Chem. 5, 1539–1541 (1966).

8. G. T. Kerr, Zeolite ZK-5: A New Molecular Sieve. Science. 140, 1412– 1412 (1963).

9. R. J. Argauer, G. R. Landolt, Crystalline zeolite zsm-5 and method of preparing the same (1972), (available at http://www.google.se/patents/US3702886).

10. R. L. Wadlinger, G. T. Kerr, E. J. Rosinski, Catalytic composition of a crystalline zeolite (1967), (available at http://www.google.com/patents/US3308069).

11. M. Moliner, C. Martínez, A. Corma, Multipore Zeolites: Synthesis and Catalytic Applications. Angew. Chem. Int. Ed. 54, 3560–3579 (2015).

12. M. Moliner, F. Rey, A. Corma, Towards the Rational Design of Effi- cient Organic Structure-Directing Agents for Zeolite Synthesis. Angew. Chem. Int. Ed. 52, 13880–13889 (2013).

13. P. Tian, Y. Wei, M. Ye, Z. Liu, Methanol to Olefins (MTO): From Fun- damentals to Commercialization. ACS Catal. 5, 1922–1938 (2015).

14. M. M. Lozinska et al., Understanding Carbon Dioxide Adsorption on Univalent Cation Forms of the Flexible Zeolite Rho at Conditions Rele- vant to Carbon Capture from Flue Gases. J. Am. Chem. Soc. 134, 17628–17642 (2012).

15. J. Shang et al., Discriminative Separation of Gases by a “Molecular Trapdoor” Mechanism in Chabazite Zeolites. J. Am. Chem. Soc. 134, 19246–19253 (2012).

16. T. Remy et al., Adsorption and Separation of CO2 on KFI Zeolites: Ef- fect of Cation Type and Si/Al Ratio on Equilibrium and Kinetic Proper- ties. Langmuir. 29, 4998–5012 (2013).

17. M. Moliner, C. Martínez, A. Corma, Synthesis Strategies for Preparing Useful Small Pore Zeolites and Zeotypes for Gas Separations and Catal- ysis. Chem. Mater. 26, 246–258 (2014).

18. A. M. Beale, F. Gao, I. Lezcano-Gonzalez, C. H. F. Peden, J. Szanyi, Recent advances in automotive catalysis for NOx emission control by small-pore microporous materials. Chem. Soc. Rev. 44, 7371–7405 (2015).

19. P. Eliášová et al., The ADOR mechanism for the synthesis of new zeolites. Chem. Soc. Rev. 44, 7177–7206 (2015).

20. E. Verheyen et al., Design of zeolite by inverse sigma transformation. Nat. Mater. 11, 1059–1064 (2012).

21. W. J. Roth et al., A family of zeolites with controlled pore size prepared using a top-down method. Nat. Chem. 5, 628–633 (2013).

22. M. Mazur et al., Synthesis of “unfeasible” zeolites. Nat. Chem. 8, 58–62 (2016).

23. W. H. Taylor, I. The structure of analcite (NaAlSi2O6· H2O. Z. Kristal- logr. 74, 1–19 (1).

24. W. H. Taylor, R. Jackson, The Structure of Edingtonite. Z. Kristallogr. 86, 53–64 (1933).

25. W. H. Taylor, C. A. Meek, W. W. Jackson, The Structures of the Fi- brous Zeolites. Z. Kristallogr. 84, 373–398 (1933).

100

26. L. Pauling, The structure of some sodium and calcium aluminosilicates. Proc. Natl. Acad. Sci. U. S. A. 16, 453–459 (1930).

27. L. Pauling, XXII. The Structure of Sodalite and Helvite. Z. Kristallogr. 74, 213–225 (1930).

28. U. Kolb, T. Gorelik, C. Kübel, M. T. Otten, D. Hubert, Towards automated diffraction tomography: Part I—Data acquisition. Ultramicrosco- py. 107, 507–513 (2007).

29. U. Kolb, T. Gorelik, M. T. Otten, Towards automated diffraction tomography. Part II—Cell parameter determination. Ultramicroscopy. 108, 763–772 (2008).

30. E. Mugnaioli, T. Gorelik, U. Kolb, “Ab initio” structure solution from electron diffraction data obtained by a combination of automated diffraction tomography and precession technique. Ultramicroscopy. 109, 758–765 (2009).

31. D. Zhang, P. Oleynikov, S. Hovmöller, X. Zou, Collecting 3D electron diffraction data by the rotation method. Z. Kristallogr. 225, 94–102 (2010).

32. W. Wan, J. Sun, J. Su, S. Hovmöller, X. Zou, Three-dimensional rotation electron diffraction: software RED for automated data collection and data processing. J. Appl. Crystallogr. 46, 1863–1873 (2013).

33. J. Jiang, Y. Yun, X. Zou, J. L. Jorda, A. Corma, ITQ-54: a multidimensional extra-large pore zeolite with 20 × 14 × 12-ring channels. Chem. Sci. 6, 480–485 (2014).

34. R. Martínez-Franco et al., Synthesis of an extra-large molecular sieve using proton sponges as organic structure-directing agents. Proc. Natl. Acad. Sci. 110, 3749–3754 (2013).

35. D. Singh et al., A complex pseudo-decagonal quasicrystal approximant, Al 37 (Co,Ni) 15.5 , solved by rotation electron diffraction. J. Appl. Crys- tallogr. 47, 215–221 (2014).

36. T. Willhammar, Y. Yun, X. Zou, Structural Determination of Ordered Porous Solids by Electron Crystallography. Adv. Funct. Mater. 24, 182– 199 (2014).

37. Q. Yao et al., Series of Highly Stable Isoreticular Lanthanide Metal– Organic Frameworks with Expanding Pore Size and Tunable Lumines- cent Properties. Chem. Mater. 27, 5332–5339 (2015).

101

38. Y. Yun et al., The first zeolite with a tri-directional extra-large 14-ring pore system derived using a phosphonium-based organic molecule. Chem. Commun. 51, 7602–7605 (2015).

39. Y. Yun et al., Phase identification and structure determination from multiphase crystalline powder samples by rotation electron diffraction. J. Appl. Crystallogr. 47, 2048–2054 (2014).

40. Y.-B. Zhang et al., Single-Crystal Structure of a Covalent Organic Framework. J. Am. Chem. Soc. 135, 16336–16339 (2013).

41. J. Jiang et al., Synthesis and Structure Determination of the Hierarchical Meso-Microporous Zeolite ITQ-43. Science. 333, 1131–1134 (2011).

42. Y. Lorgouilloux et al., IM-17: a new zeolitic material, synthesis and structure elucidation from electron diffraction ADT data and Rietveld analysis. RSC Adv. 4, 19440–19449 (2014).

43. R. Pophale, P. A. Cheeseman, M. W. Deem, A database of new zeolite- like materials. Phys. Chem. Chem. Phys. 13, 12407–12412 (2011).

44. J. Sun et al., The ITQ-37 mesoporous chiral zeolite. Nature. 458, 1154– 1157 (2009).

45. L. B. McCusker, Zeolite structures handout in Stockholm University.

46. C. Baerlocher, L. B. McCusker, SBUs in the Database of Zeolite Struc- tures:http://izasc.fos.su.se/fmi/xsl/IZA-SC/SBUList.htm.

47. Z. Liu et al., |(C4NH12)4|[M4Al12P16O64] (M = Co, Zn): New Heteroatom- Containing Aluminophosphate Molecular Sieves with Two Intersecting 8-Ring Channels. Inorg. Chem. 51, 1969–1974 (2012).

48. R. E. Morris, A. Burton, L. M. Bull, S. I. Zones, SSZ-51-A New Alumi- nophosphate Zeotype: Synthesis, Crystal Structure, NMR, and Dehy- dration Properties. Chem. Mater. 16, 2844–2851 (2004).

49. C. Baerlocher, L. B. McCusker, CBUs in the Database of Zeolite Struc- tures http://izasc.fos.su.se/fmi/xsl/IZA-SC/CBUList.htm.

50. Y. Li, X. Zou, SU-16: A Three-Dimensional Open-Framework Borogermanate with a Novel Zeolite Topology. Angew. Chem. Int. Ed. 44, 2012–2015 (2005).

51. C. Baerlocher, L. B. McCusker, SOS framework type in the Database of Zeolite Structures http://izasc.fos.su.se/fmi/xsl/IZA-SC/ftc_tm.xsl?- db=Atlas_main&-lay=tm&STC=SOS&-find.

102

52. Q. Liu et al., NaKA sorbents with high CO2-over-N2 selectivity and high capacity to adsorb CO2. Chem. Commun. 46, 4502–4504 (2010).

53. D. Xie et al., SSZ-52, a Zeolite with an 18-Layer Aluminosilicate Framework Structure Related to That of the DeNOx Catalyst Cu-SSZ- 13. J. Am. Chem. Soc. 135, 10519–10524 (2013).

54. C. Hansell, Carbon dioxide capture: Natural gas scrubs up well. Nat. Chem. 6, 656–657 (2014).

55. G. T. Kokotailo, S. L. Lawton, D. H. Olson, W. M. Meier, Structure of synthetic zeolite ZSM-5. Nature. 272, 437–438 (1978).

56. P. H. Schipper, F. G. Dwyer, P. T. Sparrell, S. Mizrahi, J. A. Herbst, in Fluid Catalytic Cracking, M. L. Occelli, Ed. (American Chemical So- ciety, Washington, DC, 1988; http://pubs.acs.org/doi/abs/10.1021/bk- 1988-0375.ch005), vol. 375, pp. 64–86.

57. J. Biswas, I. E. Maxwell, Octane enhancement in fluid catalytic cracking: I. Role of ZSM-5 addition and reactor temperature. Appl. Catal. 58, 1–18 (1990).

58. L. Qi, Y. Wei, L. Xu, Z. Liu, Reaction Behaviors and Kinetics during Induction Period of Methanol Conversion on HZSM-5 Zeolite. ACS Catal. 5, 3973–3982 (2015).

59. T. Yashima, H. Ahmad, K. Yamazaki, M. Katsuta, N. Hara, Alkylation on synthetic zeolites: I. Alkylation of toluene with methanol. J. Catal. 16, 273–280 (1970).

60. S. Cavallaro, L. Pino, P. Tsiakaras, N. Giordano, B. S. Rao, Alkylation of toluene with methanol III: Para-selectivity on modified ZSM-5 zeolites. Zeolites. 7, 408–411 (1987).

61. J. M. Thomas, J. M. Thomas, W. J. Thomas, Principles and Practice of Heterogeneous Catalysis (John Wiley & Sons, 2015).

62. C. Giacovazzo, Phasing in Crystallography: A Modern Perspective (OUP Oxford, 2013).

63. X. Zou, S. Hovmöller, P. Oleynikov, Electron Crystallography: Elec- tron Microscopy and Electron Diffraction (OUP Oxford, 2011).

64. http://reference.iucr.org/dictionary/Crystal.

65. T. Hahn, Ed., International Tables for Crystallography: Space-group symmetry (International Union of Crystallography, Chester, England, ed.

103

1, 2006; http://it.iucr.org/Ab/), vol. A of International Tables for Crys- tallography.

66. P. Coppens, in International Tables for Crystallography, U. Shmueli, Ed. (International Union of Crystallography, Chester, England, ed. 1, 2006), vol. B, pp. 10–24.

67. J. Grins, L. Banos, D. C. Sinclair, A. R. West, Synthesis and crystal structure of Li2NaTa7O19. J Mater Chem. 4, 445–447 (1994).

68. S. Hovmöller, CRISP: crystallographic image processing on a personal computer. Ultramicroscopy. 41, 121–135 (1992).

69. G. Oszlányi, A. Sütő, The charge flipping algorithm. Acta Crystallogr. A. 64, 123–134 (2008).

70. G. Oszlányi, A. Sütő, Ab initio structure solution by charge flipping. II. Use of weak reflections. Acta Crystallogr. A. 61, 147–152 (2005).

71. G. Oszlányi, A. Sütő, Ab initio structure solution by charge flipping. Acta Crystallogr. A. 60, 134–141 (2004).

72. J. Wu, K. Leinenweber, J. C. H. Spence, M. O’Keeffe, Ab initio phasing of X-ray powder diffraction patterns by charge flipping. Nat. Mater. 5, 647–652 (2006).

73. C. Baerlocher, L. B. McCusker, L. Palatinus, Charge flipping combined with histogram matching to solve complex crystal structures from powder diffraction data. Z. Kristallogr. 222, 47–53 (2009).

74. A. L. Patterson, A Fourier Series Method for the Determination of the Components of Interatomic Distances in Crystals. Phys. Rev. 46, 372– 376 (1934).

75. The Nobel Prize in Chemistry 1985, (available at http://www.nobelprize.org/nobel_prizes/chemistry/laureates/1985/).

76. D. Sayre, The squaring method: a new method for phase determination. Acta Crystallogr. 5, 60–65 (1952).

77. C. Baerlocher et al., Structure of the Polycrystalline Zeolite Catalyst IM-5 Solved by Enhanced Charge Flipping. Science. 315, 1113–1116 (2007).

78. L. McCusker, C. Baerlocher, “Zeolites” in International Tables for Crys- tallography: Volume H. Powder Diffraction. in press.

104

79. J. Christensen, P. Oleynikov, S. Hovmöller, X. D. Zou, Solving Approx- imant Structures Using a “Strong Reflections” Approach. Ferroelec- trics. 305, 273–277 (2004).

80. The Nobel Prize in Physics 1901, (available at http://www.nobelprize.org/nobel_prizes/physics/laureates/1901/).

81. The Nobel Prize in Physics 1914, (available at http://www.nobelprize.org/nobel_prizes/physics/laureates/1914/).

82. The Nobel Prize in Physics 1915, (available at http://www.nobelprize.org/nobel_prizes/physics/laureates/1915/index.ht ml).

83. P. R. Rudolf, C. Saldarriaga-Molina, A. Clearfield, Preparation and x- ray powder structure solution of a novel aluminum phosphate molecular sieve, (AlPO4)3(CH3)4NOH. J. Phys. Chem. 90, 6122–6125 (1986).

84. L. McCusker, The ab initio structure determination of Sigma-2 (a new clathrasil phase) from synchrotron powder diffraction data. J. Appl. Crystallogr. 21, 305–310 (1988).

85. R. W. Grosse-Kunstleve, L. B. McCusker, C. Baerlocher, Powder Dif- fraction Data and Crystal Chemical Information Combined in an Auto- mated Structure Determination Procedure for Zeolites. J. Appl. Crystal- logr. 30, 985–995 (1997).

86. R. W. Grosse-Kunstleve, L. B. McCusker, C. Baerlocher, Zeolite structure determination from powder diffraction data: applications of the FOCUS method. J. Appl. Crystallogr. 32, 536–542 (1999).

87. S. Smeets, L. B. McCusker, C. Baerlocher, E. Mugnaioli, U. Kolb, Us- ing FOCUS to solve zeolite structures from three-dimensional electron diffraction data. J. Appl. Crystallogr. 46, 1017–1023 (2013).

88. S. Smeets et al., SSZ-87: A Borosilicate Zeolite with Unusually Flexible 10-Ring Pore Openings. J. Am. Chem. Soc. 137, 2015–2020 (2015).

89. G. M. Sheldrick, A short history of SHELX. Acta Crystallogr. A. 64, 112–122 (2008).

90. M. C. Burla et al., SIR2011 : a new package for crystal structure determination and refinement. J. Appl. Crystallogr. 45, 357–361 (2012).

91. L. Palatinus, G. Chapuis, SUPERFLIP – a computer program for the solution of crystal structures by charge flipping in arbitrary dimensions. J. Appl. Crystallogr. 40, 786–790 (2007).

105

92. V. Petříček, M. Dušek, L. Palatinus, Crystallographic Computing Sys- tem JANA2006: General features. Z. Kristallogr. 229, 345–352 (2014).

93. V. Favre-Nicolin, R. Černý, FOX , `free objects for crystallography’: a modular approach to ab initio structure determination from powder diffraction. J. Appl. Crystallogr. 35, 734–743 (2002).

94. S. Smeets et al., High-Silica Zeolite SSZ-61 with Dumbbell-Shaped Extra-Large-Pore Channels. Angew. Chem. Int. Ed. 53, 10398–10402 (2014).

95. S. Smeets et al., SSZ-45: A High-Silica Zeolite with Small Pore Open- ings, Large Cavities, and Unusual Adsorption Properties. Chem. Mater. 26, 3909–3913 (2014).

96. T. Willhammar et al., EMM-23: A Stable High-Silica Multidimensional Zeolite with Extra-Large Trilobe-Shaped Channels. J. Am. Chem. Soc. 136, 13570–13573 (2014).

97. W. Hua et al., A Germanosilicate Structure with 11×11×12-Ring Chan- nels Solved by Electron Crystallography. Angew. Chem. Int. Ed. 53, 5868–5871 (2014).

98. J. E. Schmidt, D. Xie, T. Rea, M. E. Davis, CIT-7, a crystalline, molecular sieve with pores bounded by 8 and 10-membered rings. Chem. Sci. 6, 1728–1734 (2015).

99. P. Guo et al., A zeolite family with expanding structural complexity and embedded isoreticular structures. Nature. 524, 74–78 (2015).

100. J. Sun et al., Structure determination of the zeolite IM-5 using electron crystallography. Z. Kristallogr. 225, 77–85 (2010).

101. T. Willhammar, X. Zou, Stacking disorders in zeolites and open- frameworks – structure elucidation and analysis by electron crystallography and X-ray diffraction. Z. Kristallogr. 228, 11–27 (2012).

102. J. M. Newsam, M. M. J. Treacy, W. T. Koetsier, C. B. D. Gruyter, Structural Characterization of Zeolite Beta. Proc. R. Soc. Lond. Math. Phys. Eng. Sci. 420, 375–405 (1988).

103. J. B. Higgins et al., The framework topology of zeolite beta. Zeolites. 8, 446–452 (1988).

104. M. A. Camblor, A. Corma, P. Lightfoot, L. A. Villaescusa, P. A. Wright, Synthesis and Structure of ITQ-3, the First Pure Silica Poly-

106

morph with a Two-Dimensional System of Straight Eight-Ring Chan- nels. Angew. Chem. Int. Ed. 36, 2659–2661 (1997).

105. S. Vortmann, B. Marler, H. Gies, P. Daniels, Synthesis and crystal structure of the new borosilicate zeolite RUB-13. Microporous Mater. 4, 111–121 (1995).

106. H. M. Rietveld, A profile refinement method for nuclear and magnetic structures. J. Appl. Crystallogr. 2, 65–71 (1969).

107. B. Marler, Y. Wang, J. Song, H. Gies, Topotactic condensation of layer silicates with ferrierite-type layers forming porous tectosilicates. Dal- ton Trans. 43, 10396–10416 (2014).

108. H. Gies et al., Interlayer Expansion of the Hydrous Layer Silicate RUB-36 to a Functionalized, Microporous Framework Silicate: Crystal Structure Analysis and Physical and Chemical Characterization. Chem. Mater. 24, 1536–1545 (2012).

109. P. Wu et al., Methodology for Synthesizing Crystalline Metallosili- cates with Expanded Pore Windows Through Molecular Alkoxysilyla- tion of Zeolitic Lamellar Precursors. J. Am. Chem. Soc. 130, 8178–8187 (2008).

110. J. Ruan et al., Structural Characterization of Interlayer Expanded Zeo- lite Prepared From Ferrierite Lamellar Precursor. Chem. Mater. 21, 2904–2911 (2009).

111. P. Guo et al., Ab initio structure determination of interlayer expanded zeolites by single crystal rotation electron diffraction. Dalton Trans. 43, 10593–10601 (2014).

112. G. Cao, M. J. Shah, Aluminophosphate molecular sieve, its synthesis and use (2010), (available at http://www.google.com/patents/US7708976).

113. W. Wan, S. Hovmöller, X. Zou, Structure projection reconstruction from through-focus series of high-resolution transmission electron microscopy images. Ultramicroscopy. 115, 50–60 (2012).

114. H. G. Doherty, C. J. Plank, E. J. Rosinski, Crystalline zeolite ZSM-25 (1981), (available at http://www.google.com/patents/US4247416).

115. S. B. Hong et al., in Studies in Surface Science and Catalysis (Elsevier, 2001), vol. 135 of Zeolites and Mesoporous Materials at the dawn of the 21st century Proceedings of the 13th International Zeolite Conference, p. 186.

107

116. C. Baerlocher, L. B. McCusker, Database of Zeolite Structures: http://www.iza-structure.org/databases/.

117. D. E. . Vaughan, K. . Strohmaier, Synthesis of ECR-18 — a synthetic analog of paulingite. Microporous Mesoporous Mater. 28, 233–239 (1999).

118. V. A. Blatov, G. D. Ilyushin, A. E. Lapshin, O. Y. Golubeva, Structure and chemical composition of the new zeolite ISC-1 from the data of nanocluster modeling. Glass Phys. Chem. 36, 663–672 (2011).

119. K. Momma, F. Izumi, VESTA : a three-dimensional visualization system for electronic and structural analysis. J. Appl. Crystallogr. 41, 653– 658 (2008).

120. J. Shin et al., Angew. Chem. Int. Ed., in press, doi:10.1002/anie.201510726.

121. C. Baerlocher, A. Hepp, W. M. Meier, DLS-76: A Program for the Simulation of Crystal Structures by Geometric Refinement (Institute of crystallography and petrography - ETH, 1978).

122. S. Elomari, A. W. Burton, K. Ong, A. R. Pradhan, I. Y. Chan, Synthe- sis and Structure Solution of Zeolite SSZ-65. Chem. Mater. 19, 5485– 5492 (2007).

123. J. C. Jansen, F. J. van der Gaag, H. van Bekkum, Identification of ZSM-type and other 5-ring containing zeolites by IR spectroscopy. Zeo- lites. 4, 369–372 (1984).

124. P. Guo, W. Wan, L. McCusker, C. Baerlocher, X. Zou, On the relationship between unit cells and channel systems in high silica zeolites with the “butterfly” projection. Z. Kristallogr. 230, 301–309 (2015).

125. L. Palatinus et al., Structure refinement using precession electron diffraction tomography and dynamical diffraction: tests on experimental data. Acta Crystallogr. Sect. B. 71, 740–751 (2015).

108